My Tesla Investment Thesis 2.0: Tesla's Monopoly Potential

Introduction

Tesla's stock has been on a tear over the past six months. It's gone from a three and a half year low of $178 per share on the 3rd of June, to breaking previous all time highs and soaring over $400 for the first time ever. I think the reason for this is blatantly obvious. The thousands of people who read My Tesla Investment Thesis six months ago all decided to pour their life savings into TSLA stock and drove up the stock price. The timing just fits too perfectly. I doubt that anybody really cares about Tesla reporting one if its best quarters ever in Q3, all the good signs coming out of China surrounding Gigafactory 3, or the signs that Model Y is far ahead of schedule.

The stock price reaching new all time highs doesn't just cause hype and enthusiasm, but also doubt. More and more often I see questions on fora along the lines of "Is now a good time to buy, or should I wait for a dip?", and just yesterday somebody asked if people think TSLA is becoming a bubble, because he thought TSLA is quite expensive right now. To anybody who read My Tesla Investment Thesis it should come as no surprise that I am still very bullish on TSLA stock at these prices. I've continued to add shares and long term call options to my long position for the past two months without hesitation, and I've increased my avg. cost from $215 to $235. Most recently I added a block of shares at $360 during pre-market trading just before the run up to new all time highs. Lucky timing.

However, when I look back at My Tesla Investment Thesis there are some parts that I now think are probably not entirely accurate. For example, in the financials section I showed a fairly simplistic model for estimating the potential of Tesla Network with a profit margin of ~40%. Although the FSD software itself may very well be much higher than 40% margin (such is the nature of software), predicting 40% profit margins for operating an autonomous ride-sharing business is definitely a bit optimistic. There are some other predictions from that blog that I now think are a bit too optimistic, but there are also some things that I now believe are way too conservative, such as Tesla producing only 5M vehicles in 2030.

All in all, there are a number of things I'd like to update, adjust, and share new insights on, but I love the original post too much to change it. I also doubt that many people would enjoy reading through the same thesis again, even if it had been updated in certain sections. Furthermore, I've recently been thinking a lot more about the potential Tesla has to monopolize the transportation industry, or at least come close to doing so. I've had thoughts about this possibility for a while longer, but I used to brush it off as very distinct and highly unlikely. However, that's been changing as of late because I've begun to connect more dots, and because I've realised in what a truly desperate position the rest of the automotive industry is in.

So what I've decided to do for My Tesla Investment Thesis 2.0, is to approach it from a slightly different angle. I've decided to look at what potential Tesla has to monopolize the industries it's competing in, and how it would transform the company if it succeeds in doing so. As I'm writing this it's noon on the 23rd of December 2019, and I have exactly nine days left until the new year. Let's see if I can finish this before the fireworks bring us a fresh new 2020.

If you did not read my original Tesla Investment Thesis, I'd recommend you read that first. Especially if you don't know that much about Tesla, because that thesis covers more basic topics and explains a lot more about Tesla as a company. This one covers less basic Tesla knowledge, and covers a few more advanced topics such as how to solve autonomy.

Finally, I also recommend you grab yourself some snacks and stock up on drinks. No seriously! If you thought my thesis from six months ago was long, you've seen nothing yet. You've been warned. This blog post gives new meaning to the word enormous. Here's an overview of what we'll cover:
  1. Tesla's Most Important Competitive Advantages
  2. Master Plan, Part Deux
  3. Tesla Energy
  4. Automotive
  5. Autonomy
  6. AMaaS
  7. Non-Terranean Transportation
  8. Financials
With that out of the way, let's get started!

Tesla's Most Important Competitive Advantages

An exhaustive list of Tesla's competitive advantages is extremely long, but Tesla has three assets in particular that have contributed immensely to its success, and are fundamental to Tesla's potential to monopolize the industries it is competing in. Therefore I am going to discuss these three things first.

1: Batteries

Batteries are of vital importance to Tesla's main automotive business, as well as its energy storage business. It's also indirectly important to its future MaaS (Mobility as a Service) business, and in the future Tesla might be able to leverage its expertise in batteries to disrupt a few more industries such as boats, airplanes, and public transportation. But let's not get ahead of ourselves.

Anybody familiar with the EV industry is most likely already aware that Tesla is the world leader in battery technology. You're probably getting sick of seeing the following EV efficiency chart, but it's one good indicator of many that Tesla produces the best batteries in the world:

https://twitter.com/matty_mogul/status/1112752632395784193?lang=da

Another, perhaps better, indicator of who leads in battery technology is the cost per kWh. Unfortunately Tesla no longer publicly discloses their cost per kWh, so instead let's look at the prices of various battery storage solutions.

https://www.ecowho.com/battery/sort/name/
Ignore the "Cost per warranted kWh", because it factors in how many battery cycles per day are covered under warranty.

Going over these energy storage products one-by-one, you'll find out that the Tesla Powerwall 2 is ~50% cheaper per kWh than any of its competitors except one, the E-KwBe 5.6, which is similarly priced. The Powerwall 2 is listed on Tesla's website for $6,500, however, not $8,000 like in this chart. And I found the E-KwBe 5.6 listed for $4,500 here, so when also factoring in these data points, it appears that Tesla is able to produce batteries for much less than anybody else. This should not come as a surprise, because Tesla and its partner Panasonic have economies of scale on their side.

https://en.wikipedia.org/wiki/List_of_electric-vehicle-battery_manufacturers

All in all, I think it's hard to argue about Tesla's lead in batteries as of this writing at the end of 2019, but if you're reading this some time in mid or late 2020 after Tesla's Battery & Powertrain Investor Day, I reckon that Tesla's lead in batteries will be even larger and absolutely undeniable. During Battery & Powertrain Investor Day, Tesla is expected to announce the following:
Needles to say, a number of people are extremely excited about the details of what Tesla will announce in a few months, and it's looking highly likely that it will further cement Tesla's already multi-year lead in batteries.

What could go wrong?
The only way I could see Tesla's lead in batteries changing, is if a third party makes a massive battery breakthrough that completely levels the playing field, and if they then decided to, unlike Maxwell Technologies, not go to the world's largest consumer of batteries in the world (Tesla). Just like if you were to invent a better chopstick, you would most likely go sell it to China, if you invent a better battery, you're most likely to go sell it to Tesla. Besides, in response to a question about battery breakthroughs during a past conference call Elon mentioned that they are keeping a close eye on new battery research.

2: Elon

Elon is the reason I bought my first 80 or so TSLA shares back in 2015. When I read about how he saved both Tesla and SpaceX from the brink of death during the worst financial crisis of our time, while going through a divorce, while on the edge of personal bankruptcy, and while being publicly ridiculed by the media and his ex-wife, I decided to invest on the spot without having more than a basic understanding of Tesla's business.

After not one, but two successful internet startups (Zip2 & PayPal), Elon started Tesla and has built it into what it is today. At the beginning of this century, starting a car company was considered stupid, and starting an electric car company was considered stupidity squared, because it seemed literally impossible. Yet, today Tesla is the third most valuable car company in the world. To accomplish this, Elon has not taken a vacation in almost 20 years, has slept on the Fremont factory's floor, and has slept in Gigafactory 1. Basically, Elon does anything it takes to make his companies succeed, and succeed he does. 

https://en.wikipedia.org/wiki/Space_launch_market_competition

In my eyes it is impressive if somebody starts a company in any industry, and achieves a certain degree of success. To do so in an industry as complex, and with barriers to entry as high as the commercial launch services industry, is absolutely insane, and instantly makes you one of the most impressive human beings on earth. To not only succeed in an industry like that, but to also go from inception to essentially making all the competition obsolete in just 15 years*, is so out-of-this-world unimaginably astonishing that no words really do it justice. But objectively, that is what Elon has accomplished with SpaceX thanks in a large part to solving rocket reusability.

*And this is before Starship architecture. After Starship is regularly flying, I think all competitors can literally wrap it up.

(32:33 for the good part)

Recently it's starting to look more and more likely that similar things could happen with Tesla and that competitors just won't be able to keep up, which is what this blog is about. As long as Elon is CEO, I'm highly confident in my Tesla investment. I believe he's Tesla's biggest asset, but therefore also Tesla's biggest vulnerability.

What could go wrong?
If a CEO of a company is its biggest asset, losing that CEO would be a tremendous blow to the company. Even without Elon Tesla would still be in a great position today, but I'd definitely be a bit less optimistic about the distant future. So for Tesla's and humanity's sake, let's hope Elon lives a long and healthy life.

3: Tesla's Employees

Tesla's execution this decade has been nothing short of extraordinary.

https://hypercharts.co/tsla

https://hypercharts.co/tsla

Since the first full year of Model S production in 2013, Tesla has grown its business at a CAGR (Compounded Annual Growth Rate) of over 60%. In total, deliveries have increased from 22k to 250k, and revenue has increased from 2B to 21B. In just five years, Tesla has grown its business 10x. This would be impressive for any company, but for a company that is already worth billions in an industry as capital intensive as automotive manufacturing, that is an even crazier feat.

The reason as to why Tesla has been executing on its business plan this phenomenally, and why I think Tesla will be able to continue to execute going forward, does not lie in Elon alone, but rather in all the Tesla employees. If a company is nothing but a group of people working towards a common goal, then the talent, hard work, and dedication of that group of people are for a large part going to determine the degree to which that company will succeed. And if you are a talented, motivated, and hard working individual alive today, there is no better place to work than Tesla:
  1. Be Challenged. Tesla is notorious for being a challenging and demanding work environment. Elon is known for setting the bar extremely high, so if you're looking for a challenge, look no further than Tesla.
  2. Be Rewarded. Tesla's salaries may not be the best in the industry, but if you believe at all in what the company is doing and you factor in stock options, I think you'll be hard-pressed to find a better paying job than at Tesla over the next decade.
  3. Be Fulfilled. If you neither care about doing challenging work nor about making money, but you want to do something meaningful instead, Tesla is also your place-to-go. In 2016, Tesla was found to be the 2nd most meaningful place to work only behind SpaceX. You are literally saving the planet when you work at Tesla.
  4. Work alongside Elon Musk. Working at Tesla is a once in a lifetime opportunity to work for and alongside one of the most incredible human beings of all-time, Elon Musk.
All of this shows, because Tesla received a staggering 500,000 job applications in 2017, even though it was only able to hire 5.000 new people. It's also consistently ranked as one of the top companies people want to work for.

This doesn't just hold true for regular employees either, but also for executives. Tesla has been getting a lot of hate over the past few years for supposedly having high executive turn-over, but I have no doubt that the majority of executives who leave Tesla either do so because they don't want to work so hard, or because Elon is dissatisfied with them. Ask yourself, if you were a finance prodigy like Zach Kirkhorn, and recently became Tesla's new CFO after working your way up through the ranks over the past six years, where else are you going to do more exciting work than at Tesla, the company that is growing their revenues 10-fold over a 5-year periods, and that is on track to become the largest company in the world in the next decade? Where else are you going to make more in stock options as a result? Where else is your work going to be more meaningful? And where else are you going to work with somebody like Elon Musk? I think the same holds true for all the other executives at Tesla, such as Andrej Karpathy, Jerome Guillen, and Franz von Holzhausen. Not to say that none of them could leave for any reason whatsoever, but I don't think any of them can find more challenging and rewarding work than they are currently doing at Tesla.

One last thing I'd like to add to this section is a recent interview of Elon at the Air Force Space Pitch Day. There were a number of topics I had never heard Elon speak about before, including a lot things he has implemented at both Tesla and SpaceX to foster innovation, and to make sure employees are correctly incentivized to do the right things.


An example of the types of innovation this has lead to is the Tesla Superbottle:


This video is absolutely fascinating. In a nutshell, Tesla engineers have created one central cooling system for the Model 3, which is in stark contrast to every other car that has a number of different cooling systems for different parts of the vehicle. At traditional car manufacturers an innovation like this could never have happened, because there are too many engineering teams that are in charge of designing their own parts and their own cooling systems, and there is no cooperation between these departments. Automotive consultant Sandy Munro who teared apart and analysed the Model 3 thought "it was the best damn thing in the whole damn car".

To summarize, both Elon and the amazing people working at Tesla lead to me having a very high degree of confidence that Tesla will continue to execute like they have in the past, and that they will give competitors a run for their money.

What could go wrong?
Honestly not a whole lot. Tesla is one of the most desirable places to work, and I think this is only going to increase further as Tesla's success becomes more widespread. Even if at some point Elon is no longer involved with Tesla, I suspect it will still be a very desirable place to work, just like Apple still is.

Master Plan, Part Deux

Now that we've established Tesla's most important competitive advantages, let's take a look at what Elon is planning to use them for in the next decade. Let's take a brief look at what he's promised us through his Master Plan, Part Deux. Below is copied Tesla's Master Plan, Part Deux, with some comments from me in yellow.

Master Plan, Part Deux


Elon Musk July 20, 2016

The first master plan that I wrote 10 years ago is now in the final stages of completion. It wasn't all that complicated and basically consisted of:
  1. Create a low volume car, which would necessarily be expensive
  2. Use that money to develop a medium volume car at a lower price
  3. Use that money to create an affordable, high volume car
    And...
  4. Provide solar power. No kidding, this has literally been on our website for 10 years.
The reason we had to start off with step 1 was that it was all I could afford to do with what I made from PayPal. I thought our chances of success were so low that I didn't want to risk anyone's funds in the beginning but my own. The list of successful car company startups is short. As of 2016, the number of American car companies that haven't gone bankrupt is a grand total of two: Ford and Tesla. Starting a car company is idiotic and an electric car company is idiocy squared.
Also, a low volume car means a much smaller, simpler factory, albeit with most things done by hand. Without economies of scale, anything we built would be expensive, whether it was an economy sedan or a sports car. While at least some people would be prepared to pay a high price for a sports car, no one was going to pay $100k for an electric Honda Civic, no matter how cool it looked.
Part of the reason I wrote the first master plan was to defend against the inevitable attacks Tesla would face accusing us of just caring about making cars for rich people, implying that we felt there was a shortage of sports car companies or some other bizarre rationale. Unfortunately, the blog didn't stop countless attack articles on exactly these grounds, so it pretty much completely failed that objective.
However, the main reason was to explain how our actions fit into a larger picture, so that they would seem less random. The point of all this was, and remains, accelerating the advent of sustainable energy, so that we can imagine far into the future and life is still good. That's what "sustainable" means. It's not some silly, hippy thing -- it matters for everyone.
By definition, we must at some point achieve a sustainable energy economy or we will run out of fossil fuels to burn and civilization will collapse. Given that we must get off fossil fuels anyway and that virtually all scientists agree that dramatically increasing atmospheric and oceanic carbon levels is insane, the faster we achieve sustainability, the better.
Here is what we plan to do to make that day come sooner:
Integrate Energy Generation and Storage
Create a smoothly integrated and beautiful solar-roof-with-battery product that just works, empowering the individual as their own utility, and then scale that throughout the world. One ordering experience, one installation, one service contact, one phone app.
Alright, so the first promise from Master Plan, Part Deux, is to provide customers with an integrated solar and energy storage solution, and to then scale that throughout the world.

We can't do this well if Tesla and SolarCity are different companies, which is why we need to combine and break down the barriers inherent to being separate companies. That they are separate at all, despite similar origins and pursuit of the same overarching goal of sustainable energy, is largely an accident of history. Now that Tesla is ready to scale Powerwall and SolarCity is ready to provide highly differentiated solar, the time has come to bring them together.
Expand to Cover the Major Forms of Terrestrial Transport
Today, Tesla addresses two relatively small segments of premium sedans and SUVs. With the Model 3, a future compact SUV and a new kind of pickup truck, we plan to address most of the consumer market. A lower cost vehicle than the Model 3 is unlikely to be necessary, because of the third part of the plan described below.
So Tesla also plans to address the rest of the automotive market, mostly through the Model Y and Cybertruck.

What really matters to accelerate a sustainable future is being able to scale up production volume as quickly as possible. That is why Tesla engineering has transitioned to focus heavily on designing the machine that makes the machine -- turning the factory itself into a product. A first principles physics analysis of automotive production suggests that somewhere between a 5 to 10 fold improvement is achievable by version 3 on a roughly 2 year iteration cycle. The first Model 3 factory machine should be thought of as version 0.5, with version 1.0 probably in 2018.
Ramp up production as quickly as possible through major manufacturing improvements, check.

In addition to consumer vehicles, there are two other types of electric vehicle needed: heavy-duty trucks and high passenger-density urban transport. Both are in the early stages of development at Tesla and should be ready for unveiling next year. We believe the Tesla Semi will deliver a substantial reduction in the cost of cargo transport, while increasing safety and making it really fun to operate.
Sell not just passenger vehicles, but also cargo vehicles starting with the Tesla Semi.
Re-imagine public transportation for dense urban areas.

With the advent of autonomy, it will probably make sense to shrink the size of buses and transition the role of bus driver to that of fleet manager. Traffic congestion would improve due to increased passenger areal density by eliminating the center aisle and putting seats where there are currently entryways, and matching acceleration and braking to other vehicles, thus avoiding the inertial impedance to smooth traffic flow of traditional heavy buses. It would also take people all the way to their destination. Fixed summon buttons at existing bus stops would serve those who don't have a phone. Design accommodates wheelchairs, strollers and bikes.
I don't remember exactly when, but I know Elon commented on buses at some point during a conference call a few years ago. He said that he no longer believes that buses will be useful, because autonomy will be here soon and make them far less useful.

Autonomy
As the technology matures, all Tesla vehicles will have the hardware necessary to be fully self-driving with fail-operational capability, meaning that any given system in the car could break and your car will still drive itself safely. It is important to emphasize that refinement and validation of the software will take much longer than putting in place the cameras, radar, sonar and computing hardware.

Even once the software is highly refined and far better than the average human driver, there will still be a significant time gap, varying widely by jurisdiction, before true self-driving is approved by regulators. We expect that worldwide regulatory approval will require something on the order of 6 billion miles (10 billion km). Current fleet learning is happening at just over 3 million miles (5 million km) per day.
I should add a note here to explain why Tesla is deploying partial autonomy now, rather than waiting until some point in the future. The most important reason is that, when used correctly, it is already significantly safer than a person driving by themselves and it would therefore be morally reprehensible to delay release simply for fear of bad press or some mercantile calculation of legal liability.
According to the recently released 2015 NHTSA report, automotive fatalities increased by 8% to one death every 89 million miles. Autopilot miles will soon exceed twice that number and the system gets better every day. It would no more make sense to disable Tesla's Autopilot, as some have called for, than it would to disable autopilot in aircraft, after which our system is named.
It is also important to explain why we refer to Autopilot as "beta". This is not beta software in any normal sense of the word. Every release goes through extensive internal validation before it reaches any customers. It is called beta in order to decrease complacency and indicate that it will continue to improve (Autopilot is always off by default). Once we get to the point where Autopilot is approximately 10 times safer than the US vehicle average, the beta label will be removed.
Develop software that can drive a car autonomously 10 times safer than a human. That sounds pretty useful.

Sharing
When true self-driving is approved by regulators, it will mean that you will be able to summon your Tesla from pretty much anywhere. Once it picks you up, you will be able to sleep, read or do anything else enroute to your destination.

You will also be able to add your car to the Tesla shared fleet just by tapping a button on the Tesla phone app and have it generate income for you while you're at work or on vacation, significantly offsetting and at times potentially exceeding the monthly loan or lease cost. This dramatically lowers the true cost of ownership to the point where almost anyone could own a Tesla. Since most cars are only in use by their owner for 5% to 10% of the day, the fundamental economic utility of a true self-driving car is likely to be several times that of a car which is not.
Launch a ride-hailing service through which customers can have their cars earn money for them.

In cities where demand exceeds the supply of customer-owned cars, Tesla will operate its own fleet, ensuring you can always hail a ride from us no matter where you are.
Operate a fleet of robotaxis.

So, in short, Master Plan, Part Deux is:
Create stunning solar roofs with seamlessly integrated battery storage
Expand the electric vehicle product line to address all major segments
Develop a self-driving capability that is 10X safer than manual via massive fleet learning
Enable your car to make money for you when you aren't using it

So in summary, Tesla plans to:
  • Sell integrated solar and energy storage solutions worldwide through Tesla Energy.
  • Expand their automotive business to encompass all market segments, including cargo transport and public transportation. And rapidly scale production through massive improvements in manufacturing.
  • Develop software that can drive vehicles autonomously safer than a human can.
  • Launch a ride-hailing service and operate its own fleet of robotaxis.
But there is one more promise somewhat hidden in Master Plan, Part Deux, and it lies in the "Terrestrial Transport":
  • Develop, manufacture, and sell environmentally friendly boats and airplanes.
So now that we have an idea of where Elon wants to take Tesla, let's see if TLSA hitting $420 yesterday simply got me high on this company's potential, or if Elon truly has a chance to monopolize some of these industries with Tesla, like he's almost monopolized the commercial launch services industry with SpaceX.

Tesla Energy

When I wrote my original Tesla Investment Thesis, I wasn't very optimistic about Tesla's energy division, and I mostly wrote it off as an "I'll believe it when I see it" type of thing. However, I've come to change my mind a little bit, and I am now cautiously optimistic about Tesla Energy. It remains the part of Tesla's business I am least familiar with though, so this section will require me to do the most research and will be the most difficult to write. Therefore, let's start with the easiest part.

Energy Storage

We've already established that one of Tesla's biggest competitive advantages is the fact that it's the world leader in (lithium-ion) battery technology, in part because its energy storage products are cheaper on a $/kWh basis than any of its competitors by a large margin. And this is before further advancements are announced at next year's Battery & Powertrain Investor Day.

It's not just consumer energy storage solutions where Tesla's products are leading, but Tesla also has the potential to dominate the utility scale energy storage market. At the end of 2017 Tesla installed the world's largest battery in South Australia, a massive 100MWh project, to help prevent power outages. The battery made back a third of its cost in just one year, and the Australians have been so pleased with it that they recently got Tesla to increase the battery's size by 50%.

So if Tesla's batteries are so much better and cheaper than its competitors, then why does Tesla currently not hold a monopoly in the energy storage market? The answer lies in a lack of supply. Tesla currently simply cannot even produce enough batteries to fulfill the demand for its vehicles, and has said that it had to convert production lines that were meant for energy storage production to Model 3 battery pack production. But things are looking up a little bit for Tesla's energy storage business, because MWh deployed has once again started increasing this year.

https://hypercharts.co/tsla
(Q1'18 was unusually large because of the project in South Australia)

According to this Bloomberg forecast, the global energy storage market is expected to increase 122x from 17GWh in 2018 to 2,850 GWh in 2040, so there is an enormous business opportunity for Tesla here. Considering Tesla's lead in batteries, and presuming Tesla's world-class engineers and supply chain managers are going to be able to scale battery production rapidly, this is looking very good, but we'll have more clarity after the upcoming Battery & Powertrain Investor Day.

What could go wrong?
Assuming Tesla maintains its large lead in batteries and its employees are able to scale production, it's just a matter of time before Tesla dominates the energy storage market. However, over the course of 10-20 years, there's a higher possibility that a third party will discover a massive battery breakthrough. But as I mentioned, in all likelihood they would take this technology to the world's largest consumer aka Tesla.

Energy Generation - Solar Roof

When I woke up this morning, I had no idea about the competitive landscape of Tesla's Solar Roof, except for that I knew it'd be competing against regular roofs and regular rooftop solar. Little did I know that Tesla is not the only company who offers a roof product that can generate energy through photovoltaics. Tesla will have to go up against a number of competitors in the BIPV (Building Integrated Photovoltaics) market as it is called. Let's take a look at who these competitors are, and what Tesla's chances are of beating them.

CertainTeed's Apollo II Tile

This is the most well established BIPV product on the market. The biggest drawback is that they only sell solar tiles, and no non-solar tiles. This means that you'll either pay much more for a roof that produces more energy than you need, or you'll have to go through the hassle of only installing the Apollo II tiles on part of your roof, and covering the rest of it with regular roof tiles. If you choose to go for the latter, you'll also end up with a roof looking like this:

https://www.certainteed.com/solar/products/apollo-tile-ii/
Not terrible looking, but a bit funky.

Pricing of the Apollo II roof isn't advertised, but it's rumored to be $2.82 per watt.

Luma Solar Roof

Luma is a much smaller private company offering a Solar Roof product similar to Tesla's. Unlike CertainTeed, Luma also offers non-solar tiles, so you don't have to mix their product into a conventional roof, and you'll end up with a sleek looking monotone roof.

https://www.lumasolar.com/

The problem with this roof however, is its price. Reportedly custom made for $4.00 per watt, this roof is at the very high end even for a premium product.

RGS Powerhouse 3.0

https://news.energysage.com/rgs-energy-now-installing-dow-powerhouse-solar-shingles/

The Powerhouse is a BIPV product originally developed by a different company, but was sold to RGS Energy in 2017, and its price is $3.30 per watt. The largest problem with this product is the financial stability of the company offering it.

https://www.zacks.com/stock/chart/RGSE/fundamental/market-cap

Over the course of just 5 years, the company's market cap has dropped from 180M to a meager 8M, and it has delisted from the Nasdaq. It has exited its traditional solar business and is currently exploring "strategic alternatives". This isn't the type of company you want to be reliant on to service and repair your solar roof over the next few decades.

SunTegra Tile

https://www.suntegrasolar.com/the-suntegra-advantage/

Like most of these products, SunTegra only sells solar tiles and thus require integration with regular roof tiles. The result doesn't look what I'd call S3XY. The price is reportedly around $3.65 per watt, which is pretty expensive.

Exasun X-Roof

https://exasun.com/en/

This is in my opinion the best looking roof of the bunch. Exasun is a Dutch company that does not operate in the USA like the other companies mentioned. Details around this product are hard to find though.

There also appear to be a few small startups that jumped on the hype train when Elon first unveiled Tesla's Solar Roof, and decided to try to capture a piece of the market. By and large these don't appear to have launched any products yet, and in the case of a company called Forward, there have apparently been numerous claims of fraud against it.

Tesla Solar Roof V3

https://www.inverse.com/article/56786-tesla-solar-roof-v3-price-rumored-features-launch-date

Tesla's Solar Roof has already gone through three iterations, and I think that's why a lot of the hype has died down over the past few years since it was first unveiled in late 2016. I think it's clear that Tesla's roof is the most beautiful one on the market, and the Dutch Exasun roof is the only one that even comes close. Considering that a large part of the reason for buying a Solar Roof is aesthetics and the increase in house price, this matters quite a bit.

In terms of cost, the previous versions of Solar Roof were always very expensive. but V3 finally seems to be offered at a compelling price. Tesla advertises it on their website for $1.99 per watt, which would significantly undercut competition, but this price includes incentives, so the actual price is likely somewhat closer to the competition.

Tesla Solar Roof's Potential

Whereas in the automotive industry it is incorrectly argued by some that the incumbents have an advantage over Tesla due to their much larger size and having more resources, in the solar roof market Tesla has this advantage. Thanks to its automotive business Tesla has a very strong brand that's only going to get stronger in the coming years as the automotive business scales further. This will help Tesla to capture a large market share of the thus far largely untapped solar roof market, because customers will recognize the Tesla brand and be more likely to trust it.

Furthermore, with its energy division being less than 10% of Tesla's overall business, it is a lot easier for Tesla to invest the necessary funds in rapidly scaling Solar Roof production, if it turns out to be a viable market opportunity. But herein lies the largest uncertainty in Tesla's Solar Roof's potential. It's a new and unproven market. Tesla's Solar Roof V3 may be a lot cheaper than previous versions, and it may be price competitive with other premium roofs, it is still an expensive product and questions remain about how widespread its adoption will be.

https://electrek.co/2019/10/27/tesla-solar-roof-v3-quote-price-decrease/

https://electrek.co/2019/10/27/tesla-solar-roof-v3-quote-price-decrease/

However, I am cautiously optimistic for a number of reasons:
  • It's a brand new technology, and further cost reductions are a given. Tesla engineers will undoubtedly further improve the product, reduce costs, simplify the installation process, and of course larger economies of scale should also significantly drive down prices.
  • Elon thus far has always delivered. If he is this confident about the product, and has spent this many years and resources on it, I bet he's done his homework. Even if the economics can never be made good enough to make it a no brainer for every new house, it's unlikely that it won't find any success.
  • There don't appear to be any other companies in this market who stand much of a chance against Tesla, except for maybe the Dutch company that appears to have a good product, although their prices are unknown.
  • The roofing industry in general doesn't strike me as an industry that has seen much change or disruption in recent history. A quick Google search does suggest that improvements have been made in insulation, fireproof roofing, and solar roofs, but it doesn't sound like anybody has used a physics first approach to rethink the fundamentals of roofing as of late. It may be that roofing technology is already extremely good, but Tesla may also be able to rethink and improve how a lot of things are done.
What could go wrong?
There is more risk and uncertainty when it comes to unproven products and markets. Tesla is currently in the process of ramping up initial Solar Roof production and installations to 1,000 per week. I am optimistic about the long term, but the biggest risk I think is that it could remain a niche product only for premium houses. If Elon wants every new roof installed in twenty years to be a Solar Roof, Tesla will have to aggressively bring down prices, and offer non-premium options. Kind of like they started with the Roadster and Model S & X, and moved on to the Model 3 later.

Energy Generation - Regular Solar

I expect Tesla to focus its consumer solar business largely on Solar Roof. It may not make sense for everybody right now, but in an ideal world every new home would be outfitted with some sort of solar roof, and there would no longer be a need for regular rooftop solar, so I expect Tesla to push its own consumer business in that direction. There is still a market for commercial and utility scale solar projects though, so not all is lost.

The regular solar business is the part of Tesla's Energy business that I am the least optimistic about, because it's highly competitive, the technology is fairly commodotized, and Tesla has no apparent advantage.

https://news.energysage.com/how-much-does-the-average-solar-panel-installation-cost-in-the-u-s/

This list of the average prices of solar panel systems of various brands shows the problem. I'm assuming that Tesla's panels are of similar price to Panasonic's, because that is their partner at their Solar Gigafactory 2 in New York. As you can see, Panasonic/Tesla is somewhere in the middle of the pack and by no means the cheapest. This shouldn't come as too big of a surprise, because unlike commercialized Lithium-ion batteries which Tesla has been pioneering for over 15 years, solar cells have been around since the 1880s and have been used in space programs since the 1950s.

So what are Tesla's chances of monopolizing the commercial and utility scale solar industries? Almost non-existent in my eyes.

What could go right?
Although I don't see how this alone could possibly lead to Tesla having a monopoly, Tesla does have one advantage over the competition in the form of their battery storage solutions. Tesla will likely be able to leverage their world's best batteries to up-sell some customers into also buying their solar products, even if those are not quite as good as the competition's. Having one provider and installer is more convenient, especially for smaller sized companies.

Tesla Energy Summary

I'm now more optimistic than ever in Tesla's energy business. I've realised that Tesla should dominate the energy storage market as long as they can massively scale the production of their amazing batteries. One of many reasons to be excited about Battery & Powertrain Investor Day.

Doing more research into the Solar Roof market has made me also realise that there is significant potential here. I'll discuss the potential size of the market and financial implications later in the Financials section of this blog, but there is a lot of potential here.

Only Tesla's regular solar business is something I'm still not very optimistic about. They'll probably continue to grow their SolarCity business a bit on the B2C side, but they're cannibalizing themselves with Solar Roof. On the B2B side I just don't see much of a competitive advantage beyond perhaps being able to bundle their product with their world-class batteries in some cases.

Automotive


It's now noon on Christmas, I'm two days into writing what is shaping up to a massive post, and we've arrived at Tesla's bread and butter, its automotive business. This is the part of Tesla's business that I have far and away the most conviction about. I think that anybody who looks at Tesla's EV business objectively, should come to the conclusion that at the very minimum Tesla will be the market leader in ten years. But the current market leaders Volkswagen & Toyota only command a market share of about 10% in consumer vehicles. So even if Tesla is likely to be the market leader, can it capture a far larger market share than that of 20%, 30%, or perhaps even a majority market share of more than 50%?

Before we explore the answer to that question, perhaps it'd be a good idea to make a brief note about what I mean with all this talk about monopolies. I am aware of the theoretical definition of a monopoly:
"The exclusive possession or control of the supply of or trade in a commodity or service."
In practice this doesn't really occur, especially when you look at the world's economy as a whole. So for example, when I talk about Tesla's potential to (near-)monopolize the passenger vehicle market, I don't mean the chance that not a single other non-Tesla vehicle will ever be sold again anywhere on the planet. What I mean by that is the potential Tesla has to be the only major player in the industry, and the chance that all competitors will be reduced to minor players with relatively small market shares. Kind of like Google that doesn't literally have a monopoly on online search, but it kinda does.

One more final note. The automotive industry will completely transition to EVs, and probably sooner than most people think. If you're reading this and don't know why this is or are skeptical, please first read the beginning of my original Tesla Investment Thesis.

Passenger Vehicles

To capture a majority market share of passenger vehicles there are a few things Tesla needs. Let's start with:

1: Superior Products

We've already established that Tesla's batteries are the best in the world, and this shows in the range of their vehicles:

https://insideevs.com/reviews/344001/compare-evs/

This chart is from mid-2019, and a number of Teslas have seen their official ranges increased since then. The Model 3 SR+ now gets 250 miles from a single charge, and the Long Range version is now listed as 322 miles. One might be inclined to think that some of these cars are competitive with the lower end versions of the Model 3, especially if one accounts for price, but one would be incorrect to think so. I'd argue that all of these non-Tesla EVs are compliance cars, because they are sold at a loss before accounting for credits and regulations, and most of them are only sold in low volumes and only in markets that have particularly strict regulations (mainly EU).

https://cleantechnica.com/2019/12/13/tesla-scores-77-of-us-electric-vehicle-sales-cleantechnica-report/

Why were there only 897 Hyundai Kona EVs (an EV with longer range than the Model 3 SR+ for 75% of the price) sold in the USA from January through September of 2019? Surely if an 'inferior' product like the Model 3 sells 133,502 units, then Hyundai could make some mad money selling the Kona in the USA? It's because Hyundai loses money for each unit they sell before accounting for credits and regulations.

Why was the Mercedes EQC launch in the US delayed by a year? They say it's because the demand in Europe is so high, but it's actually because it's a compliance car.

In terms of range at a similar price point nobody can even come close to competing with Tesla, because of Tesla's humongous lead in battery technology that is only going to get bigger with the upcoming Battery & Powertrain Investor Day.

What else is important for passenger vehicles? Safety:

https://www.tesla.com/blog/model-3-lowest-probability-injury-any-vehicle-ever-tested-nhtsa?redirect=no

Most people reading this are probably already aware, but Tesla's cars are the safest in the world, and this is only going to improve further from here with the help of Tesla's Autopilot software:

https://hypercharts.co/tsla/safety
Q3'18's data points are a bit strange and could only possibly be explained by seasonality, but otherwise there are clear improvements over time.

It has to be said that the Autopilot miles are likely skewed towards highway driving which is less accident prone than other driving situations, but it's still clear that both Tesla's safety features and its Autopilot increase safety, because they are constantly getting better and increasing the gap between Non-Autopilot and Autopilot accident rates.

What about people who prefer performance over safety? Tesla is among the best, if not the best in this category as well:

https://cleantechnica.com/2019/03/03/car-best-value-for-the-money-tesla-model-3-specs-vs-30-competitors/

Furthermore, these acceleration numbers are occasionally improved through free OTA software updates, and recently also through a downloadable $2,000 performance boost package.

It's not just numbers and specs where Tesla is far ahead of the rest of the automotive industry. Thanks to their vertical integration, they're a one-stop-shop for all your EV needs and offer a great customer experience:
There are a few more things that go into the buying decisions of consumers, especially when it comes to Teslas. However, it remains to be seen how long consumers will be doing the buying of passenger vehicles, but more on that soon. I think I've made my point clear that Tesla makes superior EVs, so let's move on to the next thing that Tesla needs to capture a majority market share of passenger vehicles.

2: BAMF Manufacturing

Sorry Elon for stealing your acronym, but you are going to need this to remain ahead of the competition in terms of price, to maintain and improve margins, and most importantly to physically produce the number of vehicles you need to produce to capture a majority market share.

https://www.statista.com/statistics/200002/international-car-sales-since-1990/

If nothing were to change besides electrification, this means that Tesla would have to produce on the order of 35-40M vehicles to capture a majority market share, a 100x increase from 2019. Time for some BAMF Manufacturing.

Six months ago when I wrote my original Tesla Investment Thesis, I did not list manufacturing among Tesla's advantages on purpose. Although it was largely due to supplier screw-ups, the problematic Model 3 ramp up was still fresh in my memory, and although Elon has been talking about 'the machine that makes the machine' and about transitioning the focus of Tesla's engineering talent to improving manufacturing since 2016, I did not think those promises alone were enough. The Gigafactory 3 in Shanghai may have been starting to take shape, it was still hard to believe finished products would come out of that factory by the end of the year, and the rapid on-going construction could've been mostly down to the Chinese.

Today however, I am quite optimistic about the state of and pace of progress of Tesla's manufacturing efforts, because Gigafactory 3 is already producing two thousand cars per week on a single 10 hour shift. Furthermore, they've just about completed the on-site battery production facility, and they're already making good progress on the next phase of the factory that is looking like it might start pumping out Model Y vehicles as early as summer 2020. Looking at the weekly drone videos of the factory the pace of progress is absurd, and you can see meaningful progress being made every single week.

Just look at the first 10 seconds of this video to understand how quickly things are progressing.

These are some extremely good signs that bode very well for Tesla's ability to ramp up production on a much bigger scale over the next decade. In my blog from just two months ago about Tesla in 2020 & 2021, I said that I was skeptical about production starting at the Gigafactory 4 in Berlin in 2021 like Tesla claims it will. I now think there's a very slight possibility (maybe 5%) that Tesla has been under promising this and could start production at the end of 2020. More likely than not the deforestation of the area will take extra time, and the Germans will not be able to build as quickly as the Chinese have, but it seems highly likely that production will start in 2021.

Going back to what Elon said about manufacturing in the Master Plan, Part Deux:
A first principles physics analysis of automotive production suggests that somewhere between a 5 to 10 fold improvement is achievable by version 3 on a roughly 2 year iteration cycle. The first Model 3 factory machine should be thought of as version 0.5, with version 1.0 probably in 2018.
It sounds like Gigafactory 3 is likely version 1.0, and Gigafactory 4 could be version 2.0, more advanced, and incorporate a number of things they've learned from the Gigafactory 3 project. This seems even more plausible considering Tesla has reported that they expect the production capacity built out at Giga 3 to be more than 50% cheaper than Fremont per unit of capacity.

https://ir.tesla.com/static-files/b2218d34-fbee-4f1f-ac95-050eb29dd42f

Tesla bears have already been complaining this year that Tesla cannot possibly continue to spend as little as it has on Capex while continuing to grow, but Gigafactory 3 is proof that they can. When Tesla's Alien Dreadnought (that's what Elon calls Tesla's manufacturing efforts) Version 3 gets here and brings further improvements in manufacturing efficiency, other car manufacturers may fall even further behind in price of their EV offerings.

I don't know enough about automotive manufacturing to say for sure whether Tesla is making manufacturing breakthroughs that the world has never seen before, and if so how hard it would be for other companies to copy these. Either way, Tesla's EVs are cheaper than those of any competitor, has some of the best margins in the industry (not as good as Porsche/Ferrari/etc though), and looks like it's only going to improve these going forward. Furthermore, Tesla has shown with Gigafactory 3 that it can rapidly scale production. Now it just has to repeat that with another couple dozen factories over the next decade.

If for some reason you are still skeptical about Tesla's BAMF manufacturing and you don't want to take my word for it, then take Sandy Munro's word for it. Sandy is a veteran consultant with decades of experience in engineering and manufacturing in a number of sectors including automotive and aerospace. Sandy and his team consult for virtually every automotive company except for Tesla, so one wouldn't expect somebody like Sandy to be praising Tesla too much, because consultants help companies by pointing out and fixing their mistakes, not just flowering them with praise. But surprise surprise, Sandy is extremely impressed with most of what Tesla is doing, constantly talks about how he can't wait to tear down Tesla's upcoming Model Y to see what next-level things they've come up with, thinks Tesla is going to CRUSH it in China with the MiC (Made-in-China) Model 3s due to better margins, and in general Sandy kind of sounds like he's become a Tesla fanboy. Sandy isn't without criticism whatsoever though, as he's mentioned that he thinks that certain parts of the Model 3 body are over-designed and wasteful. Elon seemingly agreed with Sandy, because he has fired the person who was in charge of designing the Model 3's body.

If you're interested, there's a number of interviews with Sandy Munro on YouTube. They're very interesting, and you can learn a lot about how well Tesla's cars are designed, and about Tesla's manufacturing prowess.




3: Demand

The best products at the best price points should in most cases see large demand from consumers, but if I were to make the world's best EV in my basement and secretly decide to give it away for free to the first person who asks, I still don't think there would be much demand for it because nobody would know. So let's look at some hard data to see what kind of levels of demand Tesla has achieved today, and what it could achieve in the future.

https://www.goodcarbadcar.net/2019-us-midsize-luxury-car-sales-figures/#monthlysales


In the USA, the Model 3 is crushing its segment so hard that it is selling as much as all its competitors combined. It more or less has a 50% market share in its segment. The September numbers barely didn't fit in the second image, but the Model 3 sold 20,250 out of the 35,842 midsize luxury cars sold that month. In total for Q3'19, the Model 3 sold 46,850 out of 96,868 for a 48% market share.

https://cleantechnica.com/2019/12/13/tesla-scores-77-of-us-electric-vehicle-sales-cleantechnica-report/

This image that I showed before shows that out of 195,486 EVs sold so far in 2019 in the USA, the Model 3 alone sold 133,502 for a 68% market share of the EV market, and all Teslas combined sold 167,082 for an astonishing 85% market share. All of this excludes December sales numbers which is Tesla's strongest quarter, so these numbers should look even more ridiculous in a few weeks if one includes the Dec'19 sales data.

https://www.goodcarbadcar.net/2019-us-large-luxury-car-sales-figures/

The Model S and X have been having a rough year due to the discontinuation of the Short Range variants raising the base price, the falling away of the federal tax credit, and likely cannibalization from the Model 3. But the Model S which used to be outselling all its competitors combined, is still on top of its segment, and is still the best selling Large Luxury Car in the USA.

http://carsalesbase.com/european-sales-2019-q3-premium-midsized-segment/#more-55697

It might seem like the Model 3 is selling well in Europe, but not crushing its segment. However, a few things here significantly inflate the sales numbers of the top cars on this list, and don't make it a fair comparison for the Model 3:
Furthermore, the future is electric, so it might be better to compare Tesla's sales to that of EV competitors. In the not so distant future, traditional car companies will no longer be able to rely on their ICE technologies, and will have no choice but to compete with Tesla in the EV arena, because of stricter regulations and EVs reaching cost parity with ICEVs due to falling battery prices.

http://carsalesbase.com/european-sales-2019-q3-ev-phev-segments/#more-55712


https://eu-evs.com/

I think these statistics speak for themselves. Lastly, let's take a look at the two most mature EV markets in the world, Norway and The Netherlands. In Norway a majority of cars sold today is already electric, and The Netherlands is also headed in that direction.

https://insideevs.com/news/374015/car-sales-norway-plugins-september-2019/

The Model 3 is the most sold car in Norway this year, period. In terms of EVs, it's outselling the number two 2-to-1.

https://www.marklines.com/en/statistics/flash_sales/salesfig_netherlands_2019

In The Netherlands, Tesla as a brand is already starting to flirt with double digit market share percentages, even though they are effectively only selling a single model in the country. The Model S and X are rounding errors and sold ~500 units combined Jan-Nov 2019. Furthermore, these numbers don't even include what is the end of year delivery madness in NL.


As of right now, Tesla has delivered over 11,000 Model 3s in December in The Netherlands, and they will probably surpass 12,000 by the end of the year. Looking at the ~35,000 non-Tesla cars sold in November in the country, this means that in December 2019 one out of every four cars sold will be a Tesla Model 3. These are some peak numbers of course (Tesla's strongest quarter, and a tax benefit will be reduced at start of 2020), so I don't expect Tesla to command a 25% market share of the Dutch car market with just the Model 3 long term, but the numbers are insanely impressive nonetheless.

The next question is what kind of long term global demand Tesla can garner at maturity in say ten years from now. I hope that at this point you're beginning to understand why I think it's pretty much a certainty that Tesla will be the market leader in passenger vehicles a decade from now. So far Tesla has only launched a single mass market car, and that car doesn't even compete in one of the largest market segments.

https://www.best-selling-cars.com/global/2018-full-year-international-worldwide-car-sales-and-global-market-analysis/

https://www.statista.com/statistics/276506/change-in-us-car-demand-by-vehicle-type/

As you can see, the potential for the Cybertruck (pickup) is at least as big as the Model 3 (midsize), and the potential for the Model Y (crossover/SUV) is far larger. Simply releasing additional passenger vehicle models as part of "covering all major forms of terrestrial transport" should put Tesla among the market leaders, if nothing else changes. But other things will change. The rest of the industry, which is currently milking the last drops out of their soon-to-be obsolete ICE technology, will be forced to transition to EVs. If not by increasingly strict government regulations, then most definitely by dropping battery prices and further advancements in electric motor and powertrain technologies, which are relatively new compared to the ICE technologies that the automotive industry has now been perfecting for the better part of a century. Not if, but when the rest of the industry is forced to compete with Tesla in EVs and has to start selling them at a profit, because they can no longer rely on their ICE businesses to subsidize their EVs, Tesla will no longer be among the market leaders, but far ahead of anybody else. I think they even have a very good chance at a majority market share when this happens.

An argument that could be made against this is that people might not all like to drive the same car as everybody else. I think that there is truth to this, and I too would prefer to own a car that's uncommon (as long as it is still good-looking) over a car that is very common. However, if there's a large difference in the price and/or quality of the cars, I don't think this matters a whole lot. A good product will sell well, just look at how the iPhone has held onto a majority market share in a number of markets, and it's not even objectively far superior than its competitors like we've theorized Tesla's EVs will continue to be.

All in all, I think that the potential demand for Tesla's cars is likely enough to achieve a majority market share. If you're of a more skeptical nature, I don't blame you if you're not quite convinced of this yet, but to believe that the demand will not be there for Tesla to be at least a market leader in the EV future, I think you'd have to grossly misinterpret the facts and data.

4: Autonomy

Autonomy isn't something Tesla needs to capture a large market share of passenger vehicles, but rather autonomy will change the passenger vehicle market in a number of important ways once it's here, and we need to take a look at how well Tesla is positioned to deal with these.

The first of these are ADAS (Advanced Driver Assistance Systems). ADAS already exist, will continue to get better as the world gets closer to full autonomy, and play a role in consumers' buying decisions. Because Tesla is the only car manufacturer working on full autonomy and gradually releasing updates into consumer vehicles, Tesla's ADAS is far and away the most capable in the world, and is only going to continue to get better, as it can already drive 200+ miles autonomously with no disengagements. No other system available in consumer vehicles today can even come close to doing this.


There is another benefit to consumers of Tesla's strategy to gradually improve their software to the point of FSD (full self-driving). Because all Teslas sold today have all the hardware needed for FSD, if Tesla is also able to improve their software to that point, with a single OTA software update customers' cars will suddenly be able to drive themselves and significantly appreciate in value overnight. All other cars are basically guaranteed to depreciate in value over time, but Teslas have a chance to be worth more in five years than they are now.

The much more profound change that autonomy will bring along with it, is the complete change of who is going to buy passenger vehicles. As of today, the majority of passenger vehicles are sold to individual consumers for private use. However, AVs (autonomous vehicles) will be much more expensive, but at the same time drive down prices of taxi services. I will explain this change in more detail later, but the effect this will have on the marketplace is that private car ownership will diminish, and almost all passenger vehicles will be bought with the intention to be used as a robotaxi. Some individuals might still buy cars for private ownership, but not many will buy a car and not have it earn income for them as a robotaxi when cars cost tens of thousands of dollars more than today, and can earn tens of thousands of dollars per year operating as a robotaxi.

As a result of this big change in the purpose for what a passenger vehicle is bought, some things become less important factors in buying decisions and others become more important:

Less Important:
  • Styling. This means the people who buy cars will care even less if they buy the same care as everybody else. A fleet operator isn't going to care whether everybody else has the same car as them, if it is the best car.
  • Performance. It doesn't matter much what a robotaxi's top speed and acceleration are, unless perhaps in the far distant future driving rules are changed when non-FSD cars are outlawed. I doubt this happens before 2050 though.
  • Safety. Not to say it won't matter at all anymore, but sadly somebody who is looking to make money from their robotaxi driving around others, is going to care somewhat less about safety than a person who buys a car for personal use.
More Important:
  • Ruggedness/reliability. Enter the Tesla Cybertruck, Cybercar, etc. Due to how hard it is to damage this thing, it's going to be an amazing robotaxi, because less damage = less repair costs = more profit.
  • Total Cost of Ownership. If you recall the EV efficiency chart from the very beginning of this blog in the battery section, you'll remember that Teslas are the most efficient EVs in the world by quite a margin, resulting in lower fuel costs. Tesla not aiming to profit from service, should also help Tesla achieve the lowest TCOs in the business.
  • Longevity. This is why Tesla's million mile battery + powertrain + motor is going to be a MAJOR advantage for them. For a robotaxi that drives 100,000 miles per year, the difference between a regular car that lasts 200,000 miles and a Tesla that lasts 1,000,000 miles is enormous. In the AEV (Autonomous Electric Vehicle) future, this might be Tesla's single biggest advantage.
Basically, everything will become secondary to the economics of a vehicle. Especially with their soon-to-be million mile longevity, Tesla's advantages here cannot be understated. We haven't even talked about the possibility that Tesla will be the first company to solve autonomy, and their potential to monopolize FSD technology, but even without that if you just take into account everything I've talked about, I think there is significant potential for Tesla to monopolize the passenger vehicle market based off of:
  • Their seemingly insurmountable lead in batteries.
  • Their Alien Dreadnought coming online, and potentially becoming the best factory template in the world with future iterations.
  • Proof that they can already dominate the EV market segment, even though competitors aren't even making money on their vehicles yet. They already have near monopolies in some markets like the USA with 85% market share (before their strongest month December).
  • Massive advantages in economics once passenger vehicles will be bought exclusively to be used as robotaxis.
What could go wrong?
As I mentioned, I think that it's a near certainty that Tesla will end up among the market leaders at the very least. For them not gain a majority/monopoly market share, some of Volkswagen's serious attempts at EVs (ID.3 & ID.4) in the next few years HAVE TO achieve a certain level of success. I doubt they will sell as many as the Model 3 and Model Y, but these cars are supposed to be profitable mass market vehicles. If Volkswagen does not manage to sell at least hundreds of thousands of these per year, I think the entire automotive industry is done for and can pack up their bags. Even if they succeed, they will have to be relentless in keeping up with Tesla in a number of areas such as battery technology, battery supply chain, and the economics of their cars as robotaxis. If you want to read more about the challenges the automotive industry will be facing as it transitions to AEVs, I highly recommend reading my recent piece about it: The Automotive Industry's Transition to AEVs: Which car manufacturers will survive the next decade?.

Cargo Vehicles

This section is going to be very short and straightforward, because almost every reason why Tesla will be successful in passenger vehicles that I've just discussed holds true for cargo vehicles as well. And whereas the economics of passenger vehicles will not be the be-all and end-all until full autonomy is here, the economics of cargo vehicles are already the single most important aspect of the vehicle, because the majority of these end up as commercial vehicles owned by companies, and will be used to make money.

As discussed, Tesla has a number of humongous advantages here. Tesla's lead in batteries will lead to the cheapest vehicles, Tesla's vertical integration of service will lead to the cheapest and potentially also the best service, Tesla's lead in EV efficiency will lead to the least money spent on fuel, and Tesla's 1,000,000 mile battery + powertrain + motor will lead to far and away the best longevity.

The main thing that is different from passenger vehicles is the fact that Tesla has never sold a cargo vehicle before, and we have no undeniable proof that Tesla can achieve significant market share. But although they're a bit larger than and different from the vehicles Tesla has produced thus far, I doubt that this will pose too much of a problem. It's likely that there will be some initial learning curves in manufacturing with Tesla's first cargo vehicle, the Tesla Semi, but plenty of automotive companies produce both passenger and cargo vehicles without problem, including many smaller ones such as Daimler and Suzuki.

One other thing that I'll mention here is that, in the robotaxi future a lot of the robotaxi fleet will be underutilized during non-peak hours and at night. It therefore makes sense to have a robotaxi design that can also be used to transport cargo. It might be that some of the workload that cargo vehicles do today, will be taken over by robotaxis during non-peak hours. Of course a robotaxi won't be able to transport cattle or anything like that, but one should be able to design a vehicle that can transport both passengers and cargo like groceries and packages. Once again, the Tesla Cybertruck could be a nod in that direction, and could very well end up being a dual purpose vehicle.

What could go wrong?
There's a bit more uncertainty here until we see what kind of market share the Tesla Semi is able to garner in a few years. Otherwise, the industry will face the same challenges here as it will in the passenger vehicle market. Perhaps in practice the biggest thing that will stop Tesla from achieving the same success in the cargo vehicle industry, is only having so many resources available. For now Tesla is going to have its hands full electrifying the passenger vehicle industry, fully utilizing the potential of its energy division, and all the other things from the Master Plan, Part Deux.

Public Transportation

I can't remember exactly where I heard it, but if my memory is not playing tricks with me, Elon at some point mentioned that the Tesla (mini-)bus he talked about in Master Plan, Part Deux is no longer in development at Tesla. The reason he gave is that he believes that it is no longer necessary because of the advent of autonomy. When the price of a robotaxi that brings you directly from point A to point B starts to approach the price of a bus ticket, buses start to lose a lot of their value.

Another thing that would happen when cheap robotaxis are abundant, is that the demand for taxis would increase dramatically, and therefore the amount of vehicles on the road will also increase significantly. Traffic congestion is already a worldwide problem:

https://en.wikipedia.org/wiki/Traffic_congestion

So an increase in vehicles on the road is going to tax our infrastructures even further. Elon's solution to this, and the way he has re-imagined public transportation, is The Boring Company.

https://www.boringcompany.com/

Now I'm not going to argue that Tesla will monopolize public transportation because they'll eventually acquire TBC (The Boring Company) and create a network of tunnels under and in between all cities in the world. For one, that probably would optimistically take decades. Two, although it seems highly likely that cooperation between TBC and Tesla will continue and increase, it remains to be seen if Tesla will ever outright acquire TBC. Elon doesn't exactly like his companies to be public after all. Three, MRTs are pretty awesome, and I don't see them disappearing within my lifetime. Four, not counting their test tunnel, TBC has yet to complete a single real project. Five, nobody knows how financially viable, if at all, the TBC's Loop and Hyperloop will be. But Elon is Elon, so let's at least take a look at what TBC is doing, its potential, and how Elon could surprise us all over the next decade or two by transforming the world we live in once more.

Projects

Test Tunnel - Hawthorne, CA
Status: Complete

This is just a simple, short test tunnel, presumably to show off their technology to potential customers. Here is a video of a drive through:


Las Vegas Convention Center Loop - Las Vegas, NV
Status: In construction

This is a relatively small tunnel that's part of the expansion of the Las Vegas Convention Center, and it will transport people around the venue. Construction began a few months ago and Elon has claimed they will finish by the end of 2019, but it doesn't look like that'll happen. It seems likely though that it should be operational sometime in 2020. If it's a success, there are some plans to potentially expand it through Las Vegas' city center. If this happens it'll start to look a lot like other forms of public transportation:

https://www.boringcompany.com/lvcc

Dugout Loop - Los Angeles, CA
Status: In environmental review and permitting

Similar to the Las Vegas Convention Center Loop this loop will center around an event venue, The Los Angeles Dodger Stadium. The idea is to help relieve traffic congestion by connecting the stadium to a metro station, so that it becomes accessible by public transportation.

The Chicago Express Loop
Status: In contract review, environmental review and permitting

https://www.boringcompany.com/chicago/

The Chicago loop will transport passengers from the airport to downtown in futuristic fashion. It'd be a huge step for TBC if this gets successfully built, but there are some political headwinds. A new major has taken office this year, and she is not a fan of this project. Her main concern appears to be that she believes there is no way it could get built without public funds, even though Elon Musk has said he will fund it privately. Perhaps a meeting with Elon and the successful completion of the Las Vegas project could convince this new major to give the go-ahead, but for now it looks like this project is on hold.

East Coast Loop - Washington D.C. to Baltimore
Status: In environmental review and permitting

This is far and away TBC's largest project, and that's probably why it's been progressing so slowly. It was first announced all the way back in July 2017, but it's no surprise that it's taking a long time to get all the relevant agencies on board, and to get all the necessary permits when you're talking about digging a 35 mile long tunnel. I don't expect this to be completed within the next few years, but with all the potential expansion plans it is looking like it could be a really exciting way to prove TBC's Loop and Hyperloop technologies:

https://www.boringcompany.com/eastcoast/

Tesla & TBC

It is anybody's guess whether Tesla will ever acquire TBC, but even if that never happens it's likely that a successful TBC will benefit Tesla. TBC is already leveraging Tesla's EV technologies for their Loop system in which AEVs made of a modified Tesla Model X chassis will be used to transport passengers. It looks like TBC will dig and construct the tunnels, and Tesla will provide all the vehicles. TBC will also need to charge the batteries of these AEVs, so will presumably also leverage Tesla's Supercharger technology in some way.

Elon Musk's reputation should already lend TBC a large amount of credibility, but it seems like this is not the case everywhere like in Chicago. If TBC manages to successfully complete one or two projects though, this should really help them to garner interest from major cities across the world, because public transportation infrastructure is usually extremely expensive to build, but TBC promises to do it a fraction of the cost, similar to how SpaceX has drastically reduced the cost of access to space. Tesla would also benefit to a lesser extent from TBC's success.

The chances of TBC's success are hard to estimate at this point. If Elon wasn't so busy running SpaceX and Tesla (and Neuralink), I'd be confident that TBC would do some great things in the next decade based off of Elon's track-record and leadership alone. It's some really cool stuff they're doing, but even though I'm a Tesla investor with a very long investment horizon, I don't see TBC materially impacting my Tesla investment within that horizon. Assuming things go well, this looks like it has the potential to get exciting sometime in the 2030s and really transform the world in the 2040s, but for now I'm going to focus on other parts of Tesla's business.

What could go right/wrong?
Many things could go wrong. Elon's a bad-ass engineer and entrepreneur, but he is unable to dedicate a lot of time to TBC. Not to say that the people who are working on it are slouches, but they don't have Elon's track-record and startups fail more often than not. In essence, a lot of things could go wrong. On the other hand, it is entirely possible that the Las Vegas Loop will be operational in 2020, and that the city is so excited about the project that it approves the expansion plans throughout its city center. The same thing could happen in LA, and if TBC really achieves the cost reductions they are talking about, LA and LV could become a model for a new form of transportation that cities and countries around the world will be lining up for in the mid and late 2020s. Loop and Hyperloop systems in a lot of major metropolitan areas in the 2030s would be a hell of an exciting world, so let's root for the latter to happen.

Automotive Summary

This last public transportation section didn't end up being strictly automotive, but the vehicles going through the tunnels kind of are based on autos, so whatever. Anyway, we've found out that the potential of Tesla's main business is pretty damn enormous. Like I said, I think Tesla is effectively guaranteed to be the market leader if they continue to execute, and has a lot of potential to well exceed that and capture a majority market share, perhaps even (near-)monopolize this industry. What kind of financial implications this has for Tesla, and what kind of effect it could have on TSLA stock price, we'll take a look at later. First let's move on to another industry that Tesla is extremely well positioned in.

Autonomy

In the case of Tesla, its autonomy business and future ride-hailing business, Tesla Network, are intertwined and often talked about as if they are the same. Many competitors however, are aiming to compete in only one of these industries. Therefore, I'm going to talk exclusively in this section about the business that is the software and hardware that turn an ordinary vehicle into a fully self-driving vehicle. I will cover the ride-sharing business in the section after this.

Solving Autonomy

For a computer to drive a vehicle there are two key skills it needs to master: Perception and Planning. Perception is about perceiving the world around the vehicle correctly. Identifying other vehicles, pedestrians, cyclists, lane markings, traffic lights, a lot of other things, and apparently even garbage cans.

(7:26 for the garbage cans)

Planning is about using the image of the world around the vehicle from Perception, to plan how to correctly navigate through that world. In simpler words, an FSD system has to 'perceive' the world around it, and then use that information to 'plan' a route through it.

This is a slight oversimplification of an entire FSD system, because this paper explains that car navigation and car control are also part of the equation, and Nvidia's self-driving cars website hints that localization (determining the vehicle's location in the perceived world) sits in between perception and path planning.

However, car navigation is a solved problem, and we never hear about AVs (autonomous vehicles) failing because they took the wrong exit off of the highway, so this should not be holding us back from already having AVs right now. Car control does also not seem like it's holding us back. There have been great cruise control systems around for a while now, proving that controlling vehicle speed is not an issue, and there are now also a number of good lane keeping assist systems, proving that vehicle steering is also something that computers can do well. The two most difficult problems to solve are Perception and Planning.

Perception

Humans use a combination of vision and sound to perceive the world around us when we are driving. Various companies have come up with different ways for computers to perceive the world around the vehicle. Here's a brief overview of the common sensors used in the development of AVs:
  1. Cameras. These are a must. I'm not aware of anybody trying to develop AVs without cameras, and considering that our entire road transportation system is meant to be navigated with vision (traffic lights, traffic signs, etc.), I don't see how this could be possible.
  2. Radar. Radar sensors help to measure distances. They work well in adverse weather conditions, and Tesla is even using them to sense objects that cannot be seen directly, by bouncing radar beams underneath cars to sense what is happening in front of other vehicles.
  3. LIDAR. LIDAR is a little bit like radar in the sense that it detects the distance of objects. The differences are that LIDAR generally is used to detect objects in 360 degrees around the vehicle, is much more expensive, and does not work in bad weather conditions.
  4. Ultrasonics. These are similar to radar and LIDAR in that they are used to measure distances, but they tend to be cheaper and much shorter range.
The cameras are a must-have, because they are the only way to understand certain vital information such as traffic lights and traffic signs. To go from images and video footage to a representation of the world around a vehicle, companies use the science of Computer Vision. Through Machine Learning and with the help of Neural Networks we help computer to recognize objects (such as cars, pedestrians, traffic lights, etc.) with which the computer can create a virtual representation of the world around the vehicle. For the people among us that don't know how Machine Learning and Neural Networks work, a very simple explanation of how we teach a computer to recognize objects is as follows.
  1. We gather a very large amount of images of a type of object, say pedestrians, that we want to teach a computer to recognize.
  2. We translate these images into sets of numbers that the computer can understand.
  3. We create a mathematical set of computations that the computer performs on these numbers, and that will lead to a certain output of 1 in the case that the image is of a pedestrian, and 0 in the case that the image is not of a pedestrian. At first these computations, and therefore their outcomes, are completely random, and the computer will not be able to recognize pedestrians correctly.
  4. However, we then keep showing the computer images of pedestrians, and every time it is incorrect we slightly adjust the computations the computer makes on the images/numbers, so that it will correctly identify the image the next time. Exactly how to adjust the computations, is what the science of Machine Learning and Neural Networks is about.
  5. Over time, as the computer corrects the computations more and more to correct for mistakes it is making, it'll start to make less and less mistakes, because it'll have seen a large amount of images and have a good idea of what all types of pedestrians look like. You can sort of think of it as having a number of points, and a computer trying to draw a line that goes through all these points.
Kind of like happens in this video from 0:15 to 0:20 and from 1:24 to 1:30.

This is a very simple explanation of how a computer can learn to identify objects around a vehicle. When it comes to Perception for the purpose of creating an AV, we need more than just knowing what kind of objects are around the vehicle and in which direction they currently are. In many cases we also need to know the distance of these objects, and we need to understand if they are stationary or moving, and if they are moving, at what speed.

The need to detect the distance of objects around the vehicle is why there's such a diverse amount of Perception strategies deployed by different companies, and why there's a bit of a debate around whether to use LIDAR or not. Most companies rely on LIDAR to detect distances of objects around the vehicle, because it's highly accurate and extremely simple to do. The problems with LIDAR are that the sensor is very expensive (used to be tens of thousands of $s, but Waymo has developed an in-house LIDAR that supposedly costs under $10,000), and that LIDAR does not work well in adverse weather conditions. Otherwise, LIDAR sensors are a great way to detect the distances of objects.

The final pieces of the perception puzzle (detecting movement and speed of movement) can be calculated based off of difference between the detected objects, as long as the computer knows the distance of those objects. These are relatively simple as long as the detection of objects and measurements of the distance of objects are accurate.

Planning

At this point you might think that the rest is relatively simple. After all, if a computer accurately understands the world around a vehicle, it should be able to figure out how to navigate through that world, right? Wrong. If you find self-driving vehicles fascinating and have not listened to Lex Fridman's interview with George Hotz yet, I highly recommend listening to the full two hours. It's super interesting.

53:10 - 58:37

However, if you don't have that much time to spare, I recommend listening from 53:10 to 58:37 at least to hear him explain how complicated the Perception and Planning problems are, and that Perception and Planning cannot be separated from one another. It's quite complicated and technical, but what it comes down to is that the world around a vehicle is too complex for a human to describe in concrete terms. Things such as occlusion (objects hidden behind a bush, around a corner, or behind another vehicle) mean that you can't just make a list of all objects and their locations. There are also more abstract pieces of information that we humans instinctively use to drive cars, and therefore a self-driving computer's Perception layer can't simply pass on a list of objects and their directions, distances, movements, etc. to the Planning layer. In George's opinion, the Perception layer has to pass on a more abstract representation (a 1024 dimension vector in the case of his company), and the Planning layer has to learn how to use this vector filled with abstract information to learn how to drive. Any system that leaves out the more abstract information, is not 100% accurate and is missing information that is vital to safely drive a vehicle. According to him GM's Cruise and Waymo are not factoring in this abstract information in their planning systems.

Whether Perception or Planning is the more difficult problem to solve, I am not sure, but these are the two systems that need to be improved to the point where they make mistakes less often than humans do. When somebody reaches that point, they will have successfully developed a self-driving car.

One last thing I want to mention before we move on, is that there is another reason Elon is very outspoken against using LIDAR. The reason Elon thinks LIDAR is a crutch, is that most companies using LIDAR rely on it heavily, even though it's not what is going to get a system to become safer than a human driver. It only helps to measure distances, but the final 9s that have to be added to the 99.999999999% safety required to be safer than a human are going to have to be added through vision. Just think about what kind of situations you would expect a very safe, but not as safe as a human AV to trip up over. The situations that an AV will make mistakes in as it gets safer and safer, will be increasingly rare edge cases of objects and situations that are almost never encountered on the road. Cars flipped over, exotic animals on the road, and objects falling off of bridges are likely to be among these rare edge cases. The system will have to recognize first and foremost what these objects are, and for that you need a bad-ass computer vision system rather than LIDAR.

Iteration

The next step required to create an FSD system is iteration, because you need it to be safer than human drivers before you can gain regulatory approval. To be able to iterate on your system and improve it, you need two things:
  1. Large scale testing. You need a way to find out where your system makes mistakes and needs to be improved. Both the real world and simulators are used for this purpose. Initially it will be easy to find mistakes, but as a system gets better and better, you need to test at a larger and larger scale to discover new edge cases that your system trips up over.
  2. Large amounts of data. In order to fix the mistakes the system is making, you need large amounts of data of similar instances. For example, if the system made a mistake and the cause was determined to be due to the incorrect detection of a vehicle that was sliding sideways, you need a lot of examples of sideways sliding vehicles to teach your system to correctly detect sideways sliding vehicles in the future.
Another on-going debate is that of real-world data vs simulators. I'm not an FSD engineer, and although I haven't heard anyone say that simulators don't serve any purpose, I'm unsure what purpose they do serve. My guess is that it has something to do with generating supplemental data of situations that have already been encountered in the real world. Say your system messes up because of that sideways sliding vehicle, you might be able to create some extra examples of sideways sliding vehicles in a simulator.

What I do know however, is that simulators are not a substitute to real world large scale testing. The world and all possible driving situations are simply FAR too complex to simulate. On the road to full autonomy companies will come across thousands, if not millions, of edge cases that they could've never imagined, and that they could've never accurately simulated. Just like George Hotz argued that you can't make a list of things to accurately represent the world around a vehicle, you can't accurately represent everything about the world around a vehicle in a simulator. And like Elon has said, if you know in which situations your system is going to make mistakes, you don't need to simulate them to fix them.

Hardware

So now you know how to develop the software part of an FSD system, but software needs to run off of hardware, and in the case of an FSD system that hardware needs to have certain requirements. A super computer in a server room in some building may have access to all the power it requires, a computer in an AEV is going to have access to limited power if you want the vehicle to be able to travel meaningful distances. The battery capacity is limited after all.

Furthermore, all the hardware in the system (that includes sensors) will have to be economically viable. This is why Waymo started developing their own LIDARs in house to reduce costs, because the almost $100k sticker price of LIDAR systems was prohibitive to the economic feasibility of their business plan.

Regulatory Approval

So let's assume that you've gotten this far. You've solved Perception and Planning by Iterating on your system until it could drive a vehicle from point A to point B safer than a human is capable of, and you've developed hardware capable of efficiently running the system. Next, you'll want to sell this system and get mad rich, but there is one more hurdle you will have to overcome, which is regulatory approval. Nobody is going to buy your system, unless they are allowed to use it. It'd be kind of like selling medicine that is not approved by the FDA, or like selling a car that is not road legal. You can use that car on a track, and you could use an FSD system on private property, but the market for that is probably not very big.

It's hard to say exactly how this will play out, and different jurisdictions will likely handle things differently. The EU has already been giving Tesla a hard time over its semi-autonomous features. They've been slow to approve things, and they've set very strict guidelines that Tesla's software must abide by. However, when it comes to getting approval for full self-driving, one thing is for certain. Data will be king.

Imagine being a regulator and being asked to approve an FSD system. Your main concern is going to be safety, and minimizing the lives lost on the road each year. Some people also argue that it's unacceptable if people are killed by computers, but would you rather have 50.000 people die from traffic accidents in 2020 from human errors, or would you rather have 25.000 people die from traffic accidents in 2020 from computer errors? If you had to undergo a medical operation and you had to choose between a computer surgeon with a success rate of 50%, or a human surgeon with a success rate of 25%, which one would you choose? I think that as long as the increased safety of the computer can be proven beyond any reasonable doubt, people will choose the computer surgeon, and regulators will choose to approve an FSD system.

https://en.wikipedia.org/wiki/Transportation_safety_in_the_United_States

To prove beyond a doubt that your system saves lives, you'll have to prove that it will cause less fatalities than the average human in the safest countries on earth, that appear to record about 3.5 deaths per 1 billion miles. To prove this, you're going to need to not only develop a system safer than this, but you're going to have to show proof of it over tens of billions of miles. If you manage to do that, you should be able to get regulatory approval relatively easily, and you can start to sell your system.

The Players

Now that we've covered how to solve autonomy, let's take a look at all the companies trying to do so, and let's see how well they're doing.

Tesla

Tesla is one of a few companies pursuing autonomy that have chosen not use LIDAR. They realise that the entire road transportation system is designed for vision, that extremely competent vision is a must for developing an FSD system, and that everything LIDAR can do a vision only system can also learn to do. During their Autonomy Investor Day early in 2019, Tesla gave a couple of examples of how a vision only system can also be taught to accurately detect distances.


Andrej Karpathy, Tesla's Artificial Intelligence lead, talks about this from 2:16:47 to 2:22:20. He explains that just like humans are able to perceive depth from vision, computers can also learn to do so through stereo-vision, optical cues from the environment, and data annotations from Tesla's forward facing radar.

Talking about radar, Tesla's full sensor suite consists of eight cameras placed around the vehicle, a forward facing radar, and twelve ultrasonic sensors all around the vehicle. Tesla believes that it can pretty much solve autonomy with vision alone. The forward facing radar is to help detect the distances of objects in the direction in which the vehicle is travelling the fastest, and the near-field ultrasonic sensors help detect objects in very close proximity, such as a pedestrian walking by while the vehicle is backing out of a parking spot.

When it comes to hardware, Tesla has a clear measurable huge advantage over everybody else. Elon had the foresight to realise that hardware is an important piece of the puzzle back in 2016, so he poached a team of the world's best chip architects from AMD. The first chip they built exclusively for Tesla is years ahead of any other chip on the market. Nvidia unveiled a chip that is similar in capabilities to Tesla's just two weeks ago, but it is not going to be available until 2022. By that time Tesla's next generation chip will be out, which is likely going to be on a whole other level.

Far and away Tesla's biggest strength in the race to develop the first FSD system is their data, and how it allows them to iterate better than anybody else, and how it will allow them to gain regulatory approval. To help explain this to you, I'm going to quote my first Investment Thesis from six months ago:
So why do I think Tesla is going to dominate the future robotaxi market? They have a number of advantages, some of which appear to be very hard to overcome for competitors.

  1. Data. When it comes to AI and training neural networks, data is king. In terms of self-driving, the more data you have, the better your neural net will be at correctly identifying objects. The more data you have, the better your software will be at detecting other vehicle cut-ins. The more data you have, the more crazy edge cases you will be able to teach your AI about. Basically, the more data, the better.

    When it comes to data, Tesla isn't just ahead, they're not even just dominating, they're so far ahead that it seems like nobody else is even trying. Most competitors have about a hundred (or maybe in the case of Waymo a few hundred) cars on the road with expensive LIDAR equipment to test FSD software and gather data. Mobileye has their cameras in a few million cars, but these are cars sold by its customers such as BMW, so I don't believe they have any way to collect this data, and it's only from one camera, which is not enough information to support FSD. Nissan's ProPilot is used by about 350,000 cars, but like Mobileye this is only a single camera plus radar, and I am unsure if Nissan has any way to gather this data. Furthermore, Nissan only installed the camera on the cars of customers who paid for the feature. This shows a clear lack of understanding of the importance of data, and autonomy as a whole.

    Tesla, understanding the utmost importance of data when it comes to achieving full autonomy, has installed their entire sensor suite on every single car sold to customers since October of 2016, regardless of whether the customer paid for Autopilot or not. As a result, Tesla is currently gathering data from 500,000 customer cars every single day. In a year from now this will be approximately 1,000,000 cars, and in a few years they will be gathering data from millions of cars. Basically, their advantage in terms of data is only going to increase from here... exponentially.

    Data is not just paramount for creating a FSD car. Remember that I mentioned that part of launching a robotaxi service consists of convincing regulators that one's cars are safer than a human? The average miles per accident in the US is 165,000 miles, and Tesla's lead competitor Waymo is driving the second most autonomous miles. If they developed a car five times as safe as a human today, went to regulators to request regulatory approval a year from now, and showed them they only had three accidents over (let's be generous and say) 3,000,000 miles. Regulators would probably laugh in their face and say they could've gotten lucky.



    Tesla's fleet, on the other hand, drove 20 million miles per day in September of 2018. They had approximately 400-450k cars on the road back then, including some cars without the full sensor suite, but today Tesla has about 500k cars on the road with the full sensor suite, so they are gathering more than 20M miles of data every single day. Tesla is gathering more data in a single day, than all its competitors combined in an entire year, and this difference is only getting bigger.

    If another car manufacturer is smart enough, it could replicate this model, but the others simply don't stand a chance. None of the non-car manufacturer competitors can afford to buy this many cars to put on the roads and gather data. Neither can Tesla, but Tesla's customers are paying Tesla to build the cars they need to gather all the data. Simply genius. If I could only name one reason why I'm bullish on Tesla, this would be it right here.
  2. Test Drivers. But data gathering isn't the only thing that Tesla's customers are helping Tesla with in terms of autonomy. For safety reasons Tesla's competitors need to employ people to sit behind the wheel of the cars in their test fleet. Tesla's customers are literally paying Tesla to do this job.

    Part of how Tesla makes its FSD software better and better, is looking at ''Autopilot interventions". Whenever a customer is driving a Tesla in autopilot mode and notices an unsafe situation, they will take over control of the vehicle to make sure an accident doesn't happen. Every time this happens, the car sends a notification to Tesla's Autopilot Headquarters along with all the data surrounding the event. Tesla's Autopilot team then looks at what happened, and can improve the software so that it doesn't happen again. If a car manufacturer is smart, they could do the same thing, but there is no way any of the other competitors can hire half a million test drivers, whereas Tesla's customers are paying Tesla to be their test drivers. Again, simply genius.
Basically, Tesla is gathering more data (aka doing more real world testing and encountering more real world edge cases) in a single day than all its competitors in an entire year. Tesla's strategy of turning its products into its test vehicles and turning its customers into its test drivers is pretty much unbeatable. Nobody else is able to test their system on such a scale in so many different driving conditions, nobody is able to gather as much real world data to help with iteration, and nobody else has a realistic plan to gain regulatory approval except for Tesla.

Intel Mobileye

Mobileye which was acquired by Intel a few years ago, also held an Autonomy Investor Day last month (thanks to a commenter on my last blog for pointing this out).


To understand how Mobileye plans to solve the perception problem, watch from 56:10 to 1:00:20 of this video. Mobileye's CEO explains how he sees no way to create a perception system as safe as a human with current technology, so their plan is to create two separate systems to complement each other. He explains that if two systems make a mistake once every ~3,000 hours, two systems running independently will only both make a mistake once every 3,000 * 3,000 = 9,000,000 hours, and as such will be able to rival human levels of safety. Although I have no idea how they will be able to decide what to do when one system identifies a bear on the road, and the other system doesn't see anything. How do they decide which one is right? Perhaps the system would shut down and act extremely carefully until the uncertainty is resolved, but it still doesn't seem ideal.

The bigger issue that I see with this system is that Mobileye seems to assume that if they train two independent systems that make a mistake once every 3,000 hours, that there is no overlap between the mistakes the systems will make. By the way, the two systems Mobileye is creating are a vision based system like Tesla's, and a separate LIDAR system. I am highly skeptical though of their approach, and it seems extremely likely to me that the edge cases that trip up one system, will also trip up the other system.

Mobileye has a number of their own test vehicles for iteration purposes, but likely in the hundreds at best. They may be backed by Intel, but they do not have the resources to own and operate a fleet of hundreds of thousands of test vehicles like Tesla, and they are not a car manufacturer so they cannot copy Tesla's model either.

If you listen to Mobileye's presentation from 1:01:00 to 1:07:00 and listen to the CEO talk about their REM mapping initiative, you might think that Mobileye has access to similar amounts of data to train their system as Tesla. Mobileye's ADAS systems are sold to and installed in millions of vehicles of OEMs, and Mobileye uses data from the cameras in these systems to build maps. They then use these maps for the development of their AVs (and a few other business opportunities) to help with localization among other things.

High Definition Maps & Localization
Here is a short sub-section on HD Maps and localization. HD Maps to me seem like an extremely stupid and illogical idea for an FSD system. This article is the best explanation I've been able to find of what HD Maps are used for, and in essence they are used for localization aka where the vehicle is in the environment. FSD systems with HD Maps can do this by comparing the lane lines, stop lines, and other objects that it sees through perception to the ones from the map.

I could be wrong about the assumption I'm about to make, and if there is a self-driving car engineer reading this who could explain it in more detail in a comment, that'd be amazing, but it seems to me that HD Maps are a substitute to depth perception through vision. An FSD system that is already able to perceive distances of objects through computer vision, should already know the vehicle's distance to lane lines, stop lines, etc, and should be able to determine its location within the environment. It should know which lane it is in, it should know how well centered it is in the lane, and it should know how far it is away from a stop line. On the other hand, an FSD system that is unable to detect distances through vision and is relying on LIDAR for depth perception, is unable to figure out where the vehicle is located, because LIDAR can only perceive distances to 3D objects such as vehicles, pedestrians, etc, not to lane lines and stop lines. LIDAR only sees the surface that is the street.

Again, there is a chance I'm missing something here, but it really seems like HD Maps are a very illogical and unwieldy solution to the localization problem. Humans determine their position on the road through vision, and we don't have terabytes of data worth of HD Maps stored in our heads to navigate the roads with. Similar to LIDAR it kind of seems like a shortcut taken by lazy students who don't want to solve vision. At least in the case of Mobileye they have a pretty impressive system to make their maps, thanks to the millions of cars their ADAS systems are installed in.

Back to Mobileye
Anyway, Mobileye's data gathering is very different from Tesla's data gathering for a couple of reasons:
  • Mobileye does not have the ability to test their FSD systems in the vehicles of its customers' customers, and they cannot get feedback about mistakes from disengagements like Tesla.
  • The data that Mobileye is gathering is 10kb per kilometer, which is very little data. They are gathering a tiny amount of data to help them make maps, but they are not able to use any of this data to train their computer vision system.
  • Even if Mobileye wanted to, they could not train their vision systems with this fleet, because OEMs probably don't want their customers' cars uploading massive amounts of data to the cloud in the form of videos and images.
  • The vehicles in this fleet are not capable of receiving OTA updates to my knowledge, and even if some of them can, they're vehicles from dozens of different OEMs. There is no way Mobileye could update the software in these vehicles to tell them to send additional data. The software these vehicles came with only uploads the 10kb per kilometer of map data.
So in summary, Mobileye is only able to test their FSD system in a dozen or so cars that they own and operate. And most of the real world data they have access to is also from this tiny fleet.

Mobileye has some other initiatives and is developing LIDAR in-house just like Waymo. An analyst asked about this during their investor day (from 1:57:55 to 1:59:42), and Mobileye's CEO said that they believe that there's an 80% chance they don't need to do it, and could simply rely on other companies to supply them with LIDAR at a low enough price point. However, because in their eyes it is a crucial piece of the puzzle, they don't want to risk it and are developing the sensors in-house as well. If it turns out they don't need it, they will sell this sensor business at some point down the line.

Mobileye produces their own hardware chips, and can leverage a lot of Intel's expertise in that regard. Some time in 2020 they will start deploying a system with 9x EQ5 chips (1:16:00 in video), three of which are for redundancy purposes. These six chips taken together would produce a total of 144 TOPs which will be the exact same as Tesla's chip. However, Tesla's system has full redundancy, whereas I'm not sure if having 6 main chips and 3 backup chips provides the same level of safety. It also appears that Tesla's chip is more power efficient, because it requires 72 watts (1:31:50 in Tesla Autonomy Presentation) compared to 10 watt per chip for Mobileye for a total of 90 watts. All in all though, Mobileye only seems 1-2 years behind Tesla with their hardware chip, and I would not be surprised if they can leverage Intel's technology to keep up with Tesla.

Lastly, in terms of regulatory approval there is no way that Mobileye could gather enough data to convince regulators to allow their system to be used on public roads without supervision. They're barely driving millions of miles in the real world, they need to be driving tens of billions. They're off by a factor of 10,000x.

It seems like that might be the reason as to why Mobileye is taking a very different approach to government regulation (1:10:05 to 1:14:00). They are trying to establish what they call "RSS". RSS is supposed to be a reasonable standard of behavior that is acceptable on the road (e.g. cars are only allowed to break this hard), and as long as all actors in traffic behave by these rules, Mobileye will guarantee that their system never makes a mistake. For example, if a car in front of Mobileye's AV breaks very hard but within the limits, Mobileye will guarantee their AV stops in time because it'll have stuck to a safe following distance, but if the car in front of it literally comes to a dead stop that is outside of what is specified as accepted in RSS, then Mobileye's AV will not be blamed if it crashes. I haven't read through Mobileye's RSS, but I imagine it also stipulates things such as how fast cars are allowed to sway from one lane to another etc.

Mobileye seems to believe that as long as it solves Perception and has an accurate representation of the world around the vehicles, it can create a perfect Planning system that never makes a mistake as long as all actors act within the rules laid out in their RSS. They're trying to talk to regulators about this RSS, and they seem to expect to get regulatory approval this way. To me it quite frankly seems like they just don't know how else they'll ever be able to gain regulatory approval, and I doubt that this is going to be an easy sell to regulators. Seems like a Hail Mary to me.

Waymo

Waymo is the leader when it comes to 'the LIDAR approach' to autonomy. They heavily rely on LIDAR to solve Perception, and they've developed their own LIDARs in-house to drive down costs. Their LIDARs are among the best in the business for FSD purposes, and they are selling them to anybody who won't compete with their robotaxi business.

Waymo too uses HD Maps just like Mobileye, but they create their own through their own fleet, not unlike how Google created Google Maps. Waymo recently started mapping LA, but the biggest problem for Waymo is its lack of scale. Not only would it take them way longer to map the entire world, because they are using their own vehicles to create the maps instead of millions of its customers' customers' vehicles like Mobileye. The biggest weakness of HD Maps is the fact that they continuously need to be updated with any changes in the road environment, so Waymo would have to launch a very large robotaxi service in each area it maps to keep the maps up to date.

For their computer chip Waymo has been relying on Intel since 2009, which has now become one of its main competitors after Intel's acquisition of Mobileye. It seems like they've designed a custom chip exclusively for Waymo's needs.

In terms of Iteration, Waymo is in a similar spot to Mobileye. Waymo may have driven their system for over 10 billion miles autonomously, which is in the same league as Tesla, but these are all simulated miles, and as we've discussed earlier simulated miles are of limited use, and they are no substitute for real world testing. Waymo's real world testing is limited to a few hundred of its own vehicles operated by its own test drivers.

The biggest issue I see for Waymo is regulatory approval. Just like Mobileye they have no way to test their vehicles over a large enough number of miles to make any claims about achieving human levels of safety, nor about reducing traffic fatalities. There are often news articles about Waymo receiving regulatory approval to transport passengers in its AVs, but these always require safety drivers to be behind the wheel. I don't see any path for Waymo to receive regulatory approval for truly autonomous vehicles any time soon.

Others using LIDAR

The number of companies working on self-driving vehicles is enormous. This picture doesn't even do it justice, because half of these companies are just car manufacturers who have a tiny FSD research unit, and there are dozens of other smaller startups also working on the problem. Therefore I cannot cover all of them in detail, nor do I want to.

I'll say about all other companies using LIDAR that they are very unlikely to catch up to Waymo/Mobileye. LIDAR makes it possible to make a relatively well working demo in a short amount of time, and startups can leverage this to raise funding and show progress to investors, but Waymo is too far ahead in terms of vertical integration and developing its own cheap sensors, and nobody is able to create HD Maps on the scale that Mobileye is able to through the ADAS it sells to customers.

Even GM's Cruise, which is valued at about $20B, is very unlikely to leapfrog Waymo according to George Hotz.

From 1:26:04 to 1:26:44

Comma.ai

What would you get if somebody was trying to solve autonomy by copying Tesla, but started from scratch as a startup? The answer is Comma.ai. They do not use LIDAR and rely exclusively on vision, and in some cases some radar data. However, they are not able to install their system into the vehicles they sell to their customers like Tesla, because they are not a car company, so they've come up with a system that can be retrofitted to many (but not all) existing cars. Customers buy this system, which consists of a simple computer and a camera, from Comma.ai and install it into their own vehicle. They then use Comma.ai's Openpilot software which is similar to Autopilot and provides cruise control and lane keeping assist features.

Comma.ai wanted to sell both the retrofit hardware package and the software, but the company had to abandon this plan because the NHTSA demanded they comply with too many safety regulations, so they decided it would be easier to sell the hardware only and open source the software. George Hotz, the founder of Comma.ai, has said that they aim to be the Android of autonomy.

So for Perception they're using a vision based system. For Iteration they're using real world testing and data from 4,500 customer vehicles, which have so far traveled over 10 million miles. These are very good numbers compared to most of the industry, but multiple orders of magnitude away from Tesla's, and Tesla is selling way more vehicles than Comma.ai is selling retrofit hardware. To ever gain regulatory approval, assuming they can get to human safety levels, would require them to convince many more customers to buy their retrofit hardware for cruise control and lane keeping assist features.

But the biggest issues I see with Comma.ai are as follows:
  1. Their hardware chip is insufficient to reach full autonomy. I can't imagine their $600 hardware retrofit package's hardware chip is anywhere close to being powerful enough to support full autonomy.
  2. Their sensor suite is great for driver assistance features, but seems insufficient for full autonomy. It contains two windshield mounted GoPros, and I believe also leverages dashcams and radar data of cars that are equipped with them, but all of this seems insufficient. An indication to this fact is that their lane changing feature can change lanes, but cannot determine when it is safe to change lanes and requires the driver to determine this.
  3. As ADAS from companies like Mobileye become better and are included in more vehicles, the market for Comma.ai's hardware retrofit will shrink. If the open source software that runs on their hardware is not better than other ADAS, nobody will buy their hardware anymore, and they can not expand their real world testing and data gathering, that they need to improve the open source software and reach full autonomy.
  4. The company itself seems to be a little shaky. George was forced to step down as CEO, although he is still the president and lead enginner. The company also recently moved all the way from Silicon Valley to San Diego, which I don't think is a great sign either.

Pronto.ai & others not using LIDAR

There is at least one other company that has chosen not to use LIDAR, and I assume there are more. Pronto.ai's founder, Anthony Levandowski, used to be very vocal about his support for LIDAR when he worked as an engineer at Waymo. He went so far as to say "we have got to start calling Elon on his shit" in response to Elon's public stance against LIDAR, but Anthony has since then founded Pronto.ai, a startup that is very similar to Comma.ai in the way it is copying Tesla's strategy.

Pronto.ai also offers a retrofit hardware package, but instead of passenger vehicles they are targeting trucks, and their package is much more expensive at $4,999. They say this is a "deeply discounted price for a limited set of customers". It includes an unspecified number of cameras, and a radar.

Other than that, they are extremely similar to Comma.ai in their approach of offering driver assistance features (but not open source), and in their plan to slowly improve this to a fully autonomous system that's safer than a human. Even the name of their software, Copilot, is almost the same as Tesla's Autopilot and Comma.ai's Openpilot.

Other than that not much is known. We don't know how many trucks are part of their real world testing and data gathering fleet. We can assume that for $4,999 there is some sort of hardware chip included, but they are likely reliant on a supplier like Mobileye or Nvidia. All in all, their strategy seems good (a copy of Tesla), but questions remain about their progress and how big their fleet can get.

Putting it all together

If you're still with me, let's try to piece all of this autonomy information together, and let's see what Tesla's chances are in this industry. To do so, let's look at how well Tesla is positioned in each of the things important to creating an FSD system.
  • Perception. Vision is the most important element here. You can take a shortcut by using LIDAR, but you're going to need a world class computer vision system eventually to get to full autonomy. HD Maps in theory is a good alternative to vision for localization, but in practice it is extremely hard to make it work because of the scale required to make it work worldwide, and because they need to be kept up-to-date. Only Mobileye's mapping solution seems like it could work, but why not just do localization through vision. Tesla is one of a few companies entirely focused on solving vision, and among those companies Tesla is the clear leader.
  • Planning. I haven't talked much about approaches that companies have to planning, because there isn't a lot of information available. Tesla showed a number of impressive examples during their Autonomy Investor Day of how they use things like shadow mode to test and refine new features, but I have not been able to find anything about how others' approaches may differ.
  • Iteration. Real world data is everything here, and nobody has a fleet that can be used for real world testing and data gathering anywhere close to the size of Tesla's. As a matter of fact, all of Tesla's competitors combined don't even have a fleet that's in the same ballpark as Tesla's. Furthermore, Tesla's production (and therefore Tesla's fleet size) is increasing exponentially, so this is pretty much Game, Set, and Match. And I haven't even talked about things such as Operation Vacation, that once completed will supposedly automate Tesla's entire training & iteration process, so that Tesla's Autopilot team can go on a vacation, hence the name.
  • Hardware. As I've talked about, Tesla's autonomy chip is at least one to two years ahead of the competition's chips from Mobileye and Nvidia.
  • Regulatory Approval. Nobody, and I mean absolutely nobody, has a clear path to proving their system saves lives, except for Tesla. Mobileye's RSS seems like a moonshot that's unlikely to convince regulators. Tesla is the only player that will be able to gather enough data to prove beyond any reasonable doubt that their system saves lives.
Taking this all in, I think the question we have to ask ourselves is not who's going to be the first to solve autonomy, that will be Tesla for sure. The question we need to ask ourselves is whether anybody else will be able to solve autonomy before the point of no return beyond which Tesla's monopoly is inevitable. I think the answer to that second question is quite possibly "no".

Waymo does not stand a chance. Their strategy is flawed in that they rely on LIDAR and an effectively impossible to scale HD Maps system. Their fleet is also too small. Other companies using LIDAR are in even worse shape than Waymo. All the same flaws, but years behind.

The only company using LIDAR that stands a chance is Mobileye. It may be better to use vision for localization, but their HD Maps solution works at a large scale. Moreover, they are developing a vision only system in tandem to their LIDAR system. Publicly they say this is to build two complimentary systems that work together, but it could be that internally they feel like they should hedge against LIDAR being a dead-end, or that they already see the writing on the wall and have already shifted most of their focus to the vision only system. Mobileye also makes their own hardware chip that's about as close to Tesla's as can be (one to two years behind), but the crucial ingredient that Mobileye is missing is large scale real world testing and data gathering. They do not have a fleet like Tesla's, and without that regulatory approval could be a problem, and most worryingly it will take them a very long time to build a vision system powerful enough to support full autonomy. In the words of their own CEO: "I don't believe there's technology that can meet those propabilities of failure. Why do I think this? I know what the probabilities are in ADAS, and we are the leading company in providing these.".

So that leaves the companies not pursuing LIDAR. Can any of them hope to compete with Tesla? Comma.ai to be frank does not seem to have their stuff together, so I don't think they will. Pronto.ai? Maybe, but are they really going to find enough trucking companies willing to pay their 'deeply discounted price' of $4,999 per truck to build a meaningful fleet? That seems unlikely.

Therefore, in my eyes it appears very possible that Tesla will monopolize autonomy.

What could go wrong?
During their investor day Mobileye's CEO said that they believe there will be between two to four companies that dominate autonomy. If he is right and there will be a company that is able to compete with Tesla in autonomy, my money is on Mobileye. There could be advantages of HD Maps that I'm simply not aware of, and in that case Mobileye is suddenly in a much better position, because they're the clear leader in HD Mapping technology.

Another possibility is that Mobileye comes to realise the error of its ways, and decides it needs to copy Tesla's strategy ASAP. The main thing it lacks is a large fleet for large scale real world testing and data gathering. If it can somehow make deals with car manufacturers to roll out versions of its FSD software in customer vehicles, to use these vehicles for testing, to enable OTA updates, and to gather data from the vehicles to improve the Mobileye FSD system, then it'd suddenly be in a good position to compete with Tesla.

It seems unlikely that this will happen at all though, especially at a large scale in a reasonable time span. It probably takes months, if not years, to sell ADAS systems to car companies and negotiate the details of those contracts. Even if they pull all this off, Tesla would still have a large head start and infrastructural advantages like Project Dojo and Operation Vacation.

Autonomy Summary

Are you still with me? A few weeks ago I was thinking of writing an entire blog just about autonomy, and looking at how long this section has become on its own, that might've been a good idea, because this blog is starting to become a monster in size. Either way, I hope you now have a better understanding of what it takes to solve autonomy, and why Tesla is so far ahead. Their strategy is pretty much unbeatable, and in my opinion it's looking like Tesla will at the very least be the market leader in autonomy, and has very strong potential to monopolize this industry.

One final point I want to make is that Tesla's automotive and autonomy businesses are closely related. If Tesla solves autonomy tomorrow, years ahead of everyone else, they would not have to worry about demand for their vehicles for any of those years, because their vehicles would suddenly be FAR superior to anybody else's.

Similarly, Tesla's monopoly potential in its autonomy business is reliant on a strong execution in its automotive business, in particular in terms of manufacturing. If Tesla solves autonomy and continues to only sell it through its own vehicles (most likely), and if then Tesla is slow to ramp manufacturing, it might leave some room for a competitor to grab some market share. Because if Tesla's FSD system (hardware + software) is only available in Tesla vehicles, but Tesla's production is lacking, it gives competitors more time to solve autonomy and offer it for sale in non-Tesla vehicles.

AMaaS

AMaaS stands for Autonomous Mobility as a Service. It basically means autonomous transportation on demand, either for passengers who need a ride from point A to point B, or for cargo. Today, MaaS (non-autonomous AMaaS) exists in the form of ride-hailing services like Uber, Lyft, and Grabtaxi, regular taxi services, and logistics companies such as DHL. All of the transportation over roads that these services offer requires drivers, but when autonomy is solved this will slowly but surely transition towards driverless services that are able to offer the same product at a fraction of the cost. Basically, this industry is going to be disrupted BIG time, and will be unrecognizable five to ten years after autonomy is solved. Let's start off this section by looking at what will be required to operate a successful AMaaS business.

The Ingredients to Successful AMaaS

There are business models in which the AMaaS provider will not have to provide every ingredient, but all of the following are required in the overall system.
  1. Platform. This is the only ingredient that an AMaaS provider absolutely must do themselves. There has to be an interface (most likely a mobile application) ala Uber/Lyft through which passengers can order rides. To request the transportation of cargo will also require an interface or application.
  2. Vehicles. These are needed to transport the passengers/cargo.
  3. FSD System. To control the vehicles.
  4. Servicing. Vehicles will have to be serviced in the case of defects or damage.
  5. Charging. Electrification of all road vehicles is only a matter of time, so charging will be necessary to provide an AMaaS service in the future.
  6. Insurance. Even when FSD systems reach human levels of safety, they will still be far from perfect and continue to make mistakes for a while to come. If the FSD system doesn't come with insurance included, somebody else will have to take care of this.
  7. Infotainment. Not a must have, but an enormous differentiator for passenger AMaaS. If passengers have multiple choices of service providers, this could determine the winner.
Mobileye summed up the requirements for a robotaxi service slightly differently during their investor day:

https://www.youtube.com/watch?v=9JWvzuOlAKs&t=3517s

L5 is Infotainment and the Platform, L4 is also the Platform, L3 is Servicing/Charging/Insurance, L2 is Vehicles, and L1 is FSD System.

You might've noticed that Tesla does every single one of these. Tesla is even already working on a platform, but more on Tesla in a bit. Let's first look at some of the other players.

The Players

The Incumbents

There are a number of MaaS providers that might be looking to transition to AMaaS when autonomy arrives. Actually, all of the smart ones will be looking to do so, because their current business will not survive the transition from MaaS to AMaaS. A driven vehicle is simply not cost competitive with a driverless vehicle.

The incumbents in the MaaS industry are taxi companies, ride-hailing companies (Uber, Lyft, etc.), logistics companies (DHL, FedEx, trucking companies, etc.), car rental companies, and also rail and bus companies if you want to include all (under-)ground transportation, but for simplicity's sake let's leave out bus and rail.

These companies can theoretically transition to an AMaaS business model. Uber's revenue and profits (if it had any) would diminish if it started allowing AVs on its platform, but its business would be intact. Similarly, logistics companies would have to get rid of a lot of employees, but it could replace them by AVs and be competitive in the AMaaS future.

All of this relies on one important thing however, the commoditization of AVs, and herein lies the biggest danger for the incumbents. It doesn't look like the companies leading in the development of autonomy are going to allow this to happen. The license of its FSD software that Tesla is selling only allows for private use and use on their own AMaaS platform. Mobileye plans to launch its own robotaxi service in a few years, but doesn't plan to sell its FSD system to consumers until much later when prices have further decreased. This suggests that those licenses will also limit use to private use, because a vehicle with an FSD system could generate hundreds of thousands of $s operating as a robotaxi, so the cost of their LIDAR equipment should not be prohibitive to selling the system to consumers. Waymo (if they are able to solve autonomy) also appears to have its eyes set on launching its own AMaaS service, so I doubt they'd let their technology be used by competitors either.

Some companies like Lyft and Uber are developing AVs in-house, but their autonomy programs are pretty much a joke. Uber ranked dead last in disengagement statistics in California in 2018. The survival of the incumbents is 100% dependent on the commoditization of AVs, which is not looking like it'll happen.

The Automotive Industry

There are a number of car companies with mobility initiatives. They seem to realise that this is going to play a major role in the future of transportation, and they've seen the success (for now) of companies like Uber and Lyft, so they want to get in on it. However, they face the exact same issue as the MaaS incumbents in that they are 100% dependent on the commoditization of AVs.

The only tiny slimmer of hope they have of playing a role in the AMaaS industry is through a partnership with a company that solves autonomy. Tesla obviously doesn't need a car company as a partner, but companies like Mobileye and Waymo can develop all the FSD systems they want, but without actual vehicles they cannot be an AMaaS provider.

This seems difficult though, because vehicles are very commoditized, but FSD systems very much are not. Therefore the autonomy companies have all the power in the relationship. It'll be like a bar on a Friday night with one or two hot girls and a dozen guys.

Autonomy Companies

Comma.ai who only plans to sell retrofit hardware kits is the exception among the autonomy companies. Many of them have not announced clear go-to-market plans, but the leaders Mobileye and Waymo plan to leverage their autonomous technology to launch AMaaS services. To do so, Mobileye for example plans to partner with two other companies:

https://youtu.be/9JWvzuOlAKs?t=6814

The exact details of the joint venture have not been announced yet, but I'd expect that for the most part Mobileye will simply buy vehicles from Volkswagen and pay Champion Motors, an Israeli car dealer, for servicing and insurance. After launching in Israel in 2022, Mobileye plans to expand to Europe not long thereafter, and eventually other markets as well.

Waymo's plan is almost the exact same, but their rollout will be more limited due to one of the major flaws in Waymo's autonomy strategy, HD Maps.

Mobileye's and Waymo's plans will take a lot of time, and more importantly unimaginable amounts of capital to scale. Perhaps if they are able to lease vehicles that'd help a lot, but they still have to pay for the sensors and other infrastructure. Once it starts rolling and bringing in revenue though, it could potentially scale faster and faster. Except for Waymo of course, who will still have to make HD Maps of every new area they want to expand to first.

The Tesla Way

Tesla's approach is very different from anybody else's. Mobileye and Waymo's plan is fantastic in that it will allow them to take all of the profits from their AVs, but it can't be scaled fast because of the enormous capital requirements. If AVs become commoditized, an approach like Uber's has the advantage that they don't need much capital to scale, because it's an open platform that anybody can use. In return Uber will have to be satisfied with only part of the profits.

Tesla plans to combine these two approaches to get the best of both worlds. They realise that to scale fast and to capture massive market share as fast as possible, they can't just launch a service with only Tesla owned vehicles. Tesla (or any company for that matter) just doesn't have enough capital, and they also have to use their capital to scale their automotive business.

So as I'm sure many of you are already aware, what Tesla plans to do is to allow their customers to rent out their AVs on Tesla's platform, which they've called the Tesla Network, very similar to how Uber works today. This allows them to scale as fast as they can produce vehicles, which is starting to be pretty fast already and increasing at an exponential rate.

This might seem like Tesla will not be able to capture as much profit as a company that uses Mobileye's or Waymo's strategy. First of all, if one would have to choose between gaining massive market share or all of the profits, one should choose market share. Thirty percent of the profits of a worldwide AMaaS service will be worth more than 100% of the profits of a small scale localized AMaaS service.

But this strategy doesn't actually mean Tesla will have to give up a large amount of the profits. As much as Tesla has spare capital available, they can invest this into vehicles that go into their own fleet of AVs. You might think that the amount of spare capital Tesla has available is going to be a drop in the bucket, but I'll show you later in the Financials section that once Tesla Network starts to get up to speed, Tesla will have massive cash flows available that it can invest into its own fleet of AVs.

Another huge advantage that Tesla will have over everybody else in the AMaaS industry is vertical integration. Tesla has in-house every single ingredient needed to be successful in the AMaaS industry:

  1. Infotainment. I think it's no coincidence that Tesla is investing so heavily into their vehicle infotainment already. It may also help them sell cars today, and it may make them more enjoyable, but I bet the real reason is to be years ahead of the competition in infotainment when AMaaS becomes a reality.
  2. Insurance. Once again, it is no coincidence that Tesla has launched Tesla Insurance earlier this year. Yeah, it provides a slight financial benefit to their customers in the form of slightly reduced insurance premiums, but I doubt Tesla is making much, if any, profit from this. The real reason they are doing this is because of Autonomy and AMaaS.
  3. Charging. Biggest charging network on the planet? Check. Robocharger? Check.
  4. Servicing. Tesla does this in-house, and has been aggressively improving quality and efficiency as we've discussed previously.
  5. FSD System. Years and years ahead of the nearest competitor.
  6. Vehicles. Cheapest, most efficient, longest lasting EVs ever? Yup, we got 'em.
  7. Platform. Hmm, does Tesla have an AMaaS platform? As a matter of fact they've already gone through multiple versions of the Tesla Network, they just haven't launched it to the public yet! (See video below)
Watch from 8:01 to 8:54.

In terms of launch and expansion, you might expect Tesla Network to be ready worldwide with the press of a button, just like autonomy could be available worldwide with a single OTA update. I highly doubt this will happen though. I expect Tesla to launch Tesla Network in the USA first. It's likely that Tesla's FSD system will need tweaking in various geographical regions, so it seems likely that autonomy will first be ready in their biggest market, the USA, and that Tesla Network will also first be launched in their biggest market, the USA.

Another reason for this is because there will likely need to be a lot of tweaking in terms of customer experience. Tesla may have already gone through a few versions of their Tesla Network software, but its unlikely to be perfect right out of the gate. So it's very likely that they will launch on a smaller scale first to tweak and improve before expanding. Thereafter though, Tesla's rate of expansion will be very fast. They are not limited by HD Maps like Waymo, and they are also not limited by capital like Waymo and Mobileye thanks to their customer vehicles fleet.

Tesla's Unbeatable Trifecta*

Even if Tesla is not the only company to solve autonomy, even if Tesla's currently superior looking AMaaS strategy is matched by the competition, and even if Tesla's vertical integration turns out to be completely meaningless, we can come back to Tesla's automotive business. Tesla's EVs are the cheapest, most efficient, and crush everybody else's in terms of longevity. Tesla will still dominate AMaaS because they can undercut everybody on price. If your vehicles are more expensive, cost more money to operate, and last half as long, you cannot compete with Tesla.

This trifecta* that Tesla has going on with its automotive business, its autonomy business, and its AMaaS business appears to be truly unbeatable. It's looking like they have the potential to crush all three individually in a vacuum, but the world is not a vacuum. Even if one of the three doesn't work out as well as it's looking like they will, Tesla's domination in the other two will likely make up for it. Even if all three go a lot less well than it's looking like, Tesla should still find a lot of success due to the enormous synergies between these three businesses.

Going back to Tesla's three main competitive advantages. If Tesla holds onto its lead in batteries, if Elon remains CEO long enough, and if Tesla's employees keep kicking ass, I believe Tesla has serious potential to dominate, dare I say monopolize, terranean transportation. 

 *not sure if I can use the word trifecta like this, but it sounds good

What could go wrong?
Looking at AMaaS there are a few things that could cause Tesla issues. The main thing is that Tesla will have to solve autonomy first. Without autonomy there will be no AMaaS. Smaller issues that I don't expect to be too problematic will be media backlash over accidents. As long as Tesla's FSD system is safer than a human, and they can prove this with statistics, this should not ruin their plans.

Bigger issues will be brought forth by the fact that it's an entirely new business venture for Tesla. Things will not go smooth right from the beginning, and they'll likely need to iterate and improve on their AMaaS service for a while before it becomes a really great experience. The biggest issues however, are likely hard to foresee at this point. AMaaS is not just uncharted territory for Tesla, but uncharted territory for humanity. Tesla is certainly in the best position, but we have to factor in some degree of risk due to unforeseen events.

Looking at terranean transportation as a whole I honestly don't see any non-freak scenarios in which Tesla does not become the market leader at least. They do freaking everything, and they do almost all of it (service is still a work in progress) so damn well. An exciting future awaits courtesy of Tesla and Elon Musk.

Non-Terranean Transportation

Before we head into the final section covering financials, it's time for a brief breather. In Master Plan, Part Deux, Elon said he plans for Tesla to cover all major forms of terrestrial transport. Tesla's automotive business will cover all transportation over roads, and Tesla will work together with TBC to create the Loop and Hyperloop, which if successful would compete with rail and MRTs. But to cover all forms of terrestrial transportation, Tesla would also have to cover non-terranean transport over water, and through the air.

For the people skeptical about Tesla getting into this, right before the Cybertruck unveil Tesla's Instagram page briefly said Cars, Boats, & Airplanes. Take that as you will.

Boats

https://inhabitat.com/planetsolar-the-worlds-largest-solar-powered-boat-docks-in-hong-kong/

The history of electric boats is very similar to that of electric cars. When motor boats first came around electricity was actually the main form of propulsion. It was not until the 1920s, when the internal combustion engine became dominant, that it shifted to the gasoline boats that we have today.

There are many different types of boats. Some can be electrified with today's technology, and countries like Norway and China are already pushing hard to electrify the boats that can be electrified. These are boats such as ferries, and container ships that travel very short distances like the Yara Birkeland.

Similarly, it should not be difficult to bring to market electric versions of most types of pleasure boats, and in fact some pleasure water transportation devices such as jet skis are already getting electrified.

However, pure electric versions of longer distance vessels like the huge containers ships that roam our oceans, are probably not going to arrive any time soon. Battery technology has simply not advanced enough yet for these to be feasible, even with the use of solar panels.

I'd imagine that if the world were to really push for reducing carbon emissions of ships, and if Elon really wanted to do something about it, he could probably electrify inland ships with the help of charging infrastructure built out along popular shipping routes. But for the massive long distance container ships to become electric, we're going to need some huge energy storage breakthroughs.

When Tesla starts getting into the boat industry, a lot of what makes them successful in the automotive industry will also help them in the boat industry. Tesla has a massive lead in batteries, electric powertrains, and electric motors, so at the very minimum they should be able to supply these to boat companies. If boat companies are willing to electrify their offering, I could see Tesla end up simply as a supplier to them. I don't know if Tesla will want to actually manufacture boats in-house, but you never know considering how unafraid they are of vertical integration, and doing things themselves.

I'm sure Tesla could get some smart engineers together in a room, figure out how to build amazing boats based on their EV technology, and successfully compete in the industry. Whether they actually will, and if so when, is anybody's guess.

Airplanes

https://evtol.news/aircraft/grug-group-business-evtol-jet/

Although I'm not super confident Tesla will ever do more than be a supplier in the boat industry, I am super confident that Tesla will get into the airplane industry at some point. To explain to you why, I'm going to be lazy and quote myself from six months ago:
One of the improvements that Maxwell has made in battery tech is an improvement in battery energy density. Tesla's current batteries appear to have an energy density of 250 Wh/kg. Maxwell has demonstrated energy densities of over 300 Wh/kg, and has identified a path to a battery energy density of >500 Wh/kg. This means that if all goes according to plan, in the future the same battery will be able to hold more than twice as much energy.

When I first heard about this, I instantly thought of Elon's plans to design an EVTOL (Electric Verticle Take Off and Landing) Aircraft. Elon has talked about having a design for such an aircraft numerous times, and said he would love to build it, but that he believes battery energy density has to improve to over 400-500 Wh/kg to make it work.

It's definitely a long term thing, but I think that if Tesla succeeds in their automotive business, the chance that they announce plans to produce an electric aircraft by 2035 might be as high as 80-90%. After all, their CEO also just happens to be the lead designer at the world's number one rocket company, so he knows a thing or two about aerospace engineering.
If Tesla's automotive, autonomy, and AMaaS business form a trifecta that lead to Tesla dominating terranean transportation, Tesla's EV technology and Elon's expertise in aerospace engineering form a bifecta* that will lead  to Tesla dominating the airplane industry.

*I'm just making up words here

I'm not saying that as soon as Maxwell's technology allows Tesla to commercialize 500 Wh/kg batteries, they'll be selling EVTOLs en masse the next day. But I am saying that if I were to start an electric aircraft startup, I would be looking for the world's best battery, best electric motor, best electric powertrain, and best aerospace engineer. Tesla has all of these things.

What could go wrong?
As long as they hold on to the their large technological lead in EVs, they should be a major player in the boat and airplane industry, at least as a supplier. Therefore, the main thing that could go wrong would be Tesla losing this large technological lead somehow, and others catching up to them in terms of battery, electric motor, and electric powertrain technology. Also, for the airplane industry specifically, Elon ever leaving Tesla would be a very large blow to their potential in this industry.

Financials

It's now the morning of the 29th of December, and I've been pretty much writing non-stop (12 hours per day) for 6 days straight. I was very worried for most of the week that I wouldn't be able to finish this gargantuan thesis before the end of the year as I had originally set out to, because especially the autonomy section turned into a much bigger ordeal than I had originally anticipated. At this point though, I'm fairly confident that I will finish everything before the end of the year. This financials section is the last section, and afterwards I'll just have to go back over everything to check for errors and polish.

I think overall this Investment Thesis 2.0 has turned out to be more technical than the original one, but I'm going to try to make the financials section of this one a little more welcoming to readers who don't 'speak spreadsheet'. I'm going to explain much more about how this all works, and basically give you a crash course in financial modeling.

I'm going to be excluding a few things from the financials, namely Boats, Airplanes, and Public Transportation. The timelines and the details of these potential future business ventures are simply too vague at this point. I could research some numbers about the global airplane industry, make a ton of assumptions, and then create a financial model of it, but it'd end up being mostly speculation, and I don't think it's really worth my or your time getting into that. I've also excluded "Services and other" revenues and profits from this model, because Tesla does not aim to profit from it, so long term it should not impact valuation.

I'm going to cover each business (Energy, Automotive, Autonomy, AMaaS) individually first, and I will put it all together at the end. Unlike last time, I'm not going to present a bear and a bull model. The bear model I presented last time was quite frankly far too conservative, except for maybe a few points (autonomy margins, OPEX (operating expenses), valuation multiple). The single model I'll be presenting here is bullish, but by no means overly so. It assumes Tesla will continue executing similarly to how it has in the past, which I think is a reasonable assumption to make. It's possible that if something happens to Elon, these numbers to turn out to be incorrect. However, based on all the information we have available to us today, I believe this model to be on the aggressive side, but reasonable.

The model I'm presenting projects the next decade and goes up till 2030, but this wouldn't be a "Monopoly Potential" blog without financials that show Tesla's potential. So in addition to these, I have also modeled out a few different 'end-states' of what Tesla's business would look like financially if it ends up with various amounts of market share.

If you're on a computer and would like to follow along more closely with what we're doing, or if you completely disagree with things I say and think you know better, I suggest you open the financial model that I've uploaded to Google Drive below. You'll be able to see how everything works, learn about finance and financial modeling, and you can change things you disagree with me on. There are also a few parts in the model for you to fill in your own predictions, and you should be able to do so in Google Sheets with the Excel version of the model.

Tesla Investment Thesis 2.0 Financial Model - Excel
Tesla Investment Thesis 2.0 Financial Model - Numbers
Tesla Investment Thesis 2.0 Financial Model - PDF

But now without further ado, let's get into the first part, Tesla Energy.

Notes:
  • Fields in the color Cyan are guesses, assumptions, estimations, etc.
  • Fields in the color Red indicate funky data. Often due to changes in the way Tesla has reported certain numbers over time.
  • There are a few fields in the color Green. These are my estimations for Tesla's potential.
  • There are a few fields in the color Yellow. These are for you to fill in your own estimations about Tesla's potential.

Energy

For simplicity's sake, I'm going to assume all of Tesla's solar is sold, and Tesla does not lease any solar. This is of course incorrect, and I do have a model that separates between solar sales and leases, but Tesla only very recently started releasing more detailed information in regards to this, so the model is forced to make a lot of assumptions, and I believe it would just confuse most of you reading this.  Modeling all solar as direct sales instead of leases inflates revenues and profits in the short term, but underestimates them in the long term. When you lease something, you pay extra for the financial service that is being provided after all, and Tesla's automotive leasing business is a prime example of how this works.

Energy Storage

Let's start by modeling Tesla's Energy Storage business. To do this we need to figure out the following numbers under "Storage" in the "Energy" tab:


Past Data
  • MWh Deployed is always reported by Tesla. The numbers for 2016 and onwards can easily be found in Tesla's quarterly financial reports, as well as in Tesla's quarterly and annual SEC filings. The numbers for 2014 and 2015 were calculated off of $ / MWh.
  • Revenue is a little trickier. The numbers for energy generation and storage can be found in the quarterly reports, but to find the energy storage only numbers, one has to dig through the SEC filings. Even then, only the numbers for 2014, 2015, and 2016 can be found. Since 2017 Tesla has not reported separate numbers for energy storage and generation. Therefore we'll have to calculate those from the MWh deployed data and from $ / MWh estimations.
  • $ / MWh is even trickier. The only year from which we have the data on both MWh deployed and Revenue is 2016, so we can calculate that it was $993 (thousand) dollars in that year. 2014 and 2015 don't matter tremendously, but I'll guess it was around $1000. My guesses of $700 and $500 for 2017 and 2018 are mostly based on the fact that Tesla's Powerwall 2 costs $480 (thousand) dollars per MWh today.
  • Profit is almost the same as revenue. We can calculate for the years 2014 and 2015 off of Tesla's data, but we have to guess for 2016, 2017, and 2018, because no separate data is available for storage and generation.
  • Margin is what we use to guess profits. Margin we can simply calculate 2014's and 2015's from the available revenue and profit numbers, but we'll guess margin for the following years based off of Tesla's overall energy margin.

Future Projections
  • MWh Deployed I am guessing will not grow as aggressively for the next few years until 2022 as Elon and Tesla claim. They've claimed they will likely double energy storage deployments for the foreseeable future, but as of right now I'm worried about battery supply. I'm hopeful though that with announcements made at Battery & Powertrain Investor Day, Tesla will be able to scale battery production more rapidly early in the next decade. Therefore, I have MWh Deployed growing much faster after 2023. A total of 405 GWh deployed in 2030 may look like a stupidly large number, but the energy storage business is projected to grow rapidly to nearly 3,000 GWh worth $620B by 2040. 405 GWh in 2030 would be a large market share, but Tesla produces the world's best batteries, so I don't think this is too unrealistic.
  • $ / Mwh is guaranteed to drop as battery prices drop. There are a lot of different projections for how much battery prices will drop. This article claims 52% by 2030 and another claims a slightly less than 50% decrease in costs, so I went with a 50% drop in $ / MWh by 2030.
  • Margin of 20% is approximately where Tesla's energy business as a whole is at today. It's very possible Tesla could increase margins to 25%, or even 30% if they really wanted to. They're enormously far ahead of the competition after all when it comes to batteries. 30% seems unlikely though given Tesla's goal of saving the planet, but I don't blame you if you want to change these numbers to 22% or 25%. I've kept it more conservative at 20%.
  • Revenue and Profit are simply calculated from the above numbers.

Potential

You can find these under the "Energy Potential" tab. In the "Potential" sections you'll encounter green and yellow fields. The yellow fields are for you to fill in your projection, the green fields are my projections. For me these are very much based on feel, and not exactly what I know will happen. There are places where I want to be more conservative, like I might see more potential but feel like there's too much uncertainty so I'll err on the side of being conservative. They also change over time based on both good and bad news that comes out.

The Low, Medium, High, and Monopoly Potential columns are mostly for illustration purposes, and to help you see the kinds of results different kinds of numbers lead to.

Okay, let's continue. To estimate Tesla's total potential in the energy storage market in MWh Deployed per year, we need to estimate the total addressable worldwide market, and we need to think about what Tesla's market share can be.

  • Market in MWh for energy storage in 2040 is projected to be about 3,000 GWh by BloombergNEF. I haven't been able to find any other estimates, so we'll go with this one.
  • Tesla Market Share % estimation will be different for each person. If you have the spreadsheet open, you can fill in yours in the yellow field. I think Tesla's lead in batteries is undeniable and unlikely to change, so in theory they should do better than 50% market share, but it's still early and questions around their raw materials and cell supply remain in the short, mid, and maybe even long term. A market share of 50% for 1,500 TWh is already enormous, and they need batteries for their automotive business too.
  • $ / MWh I dropped this a little bit further from the previous section. Of course this could drop further by 2040 and 2050 if you're thinking super long term.
  • Margin I have kept the same as in the previous section.
Okay, let's move on to Solar.

Solar

For Solar we need very similar data, but instead of MWh Deployed we're looking for MW Deployed, and instead of $ / MWh we're looking for $ / MW.

Past Data
  • MW Deployed was reported by Tesla in their financials for 2017 and 2018. There was some non-storage energy revenue in 2016 due to the SolarCity acquisition in November of 2016, but the MW Deployed is a guess based off of $ / MW and Revenue.
  • Revenue was reported for 2016 only in this SEC filing. For 2017 and 2018 only total energy revenues were reported, so this was calculated based off of the energy storage calculation we did before (total revenue - storage revenue = generation revenue).
  • $ / MW is calculated from MW Deployed and Revenue, but this data is untrustworthy because it does not account properly for solar leasing, hence the red field color.
  • Profit is the exact same as revenue.
  • Margin flows forth from revenue and profit.

Future Projections
  • MW Deployed is something I am very conservative on. You might remember that I'm not too optimistic about Tesla Solar from a fundamental standpoint, and I think they'll be pushing their residential customers heavily towards Solar Roof in the coming years.
  • $ / MW across the industry should drop as prices come down. This research estimates prices have dropped by 80% in the last decade, and will drop by an additional 50% by 2030, so that's what I went with.
  • Margin I have lower than storage at 15%, because I think solar technology is more commodotized and Tesla does not appear to have a technological lead. They'll likely have to operate at lower margins to be competitive at all.

Potential
  • Market in MW. This article has solar capacity growing to a total of ~4,500 GW in 2040. Solar would then make up 15% of the world's electricity generation. To achieve that, it projects Solar capacity growth from ~2,500 GW to ~4,500 GW from 2030 to 2040, for a growth of about 200 GW per year. This article projects about double the annual solar generation by 2040 though, so I've gone with 250 GW installed per year. In 2017, the world installed about 100 GW of additional solar capacity.
  • Tesla Market Share % is very low for me. I think the 10% in the high potential case is quite optimistic. I just don't think Tesla is in a good position to capture more than that, and I doubt anybody will considering the technology seems quite commodotized.
  • $ / MW I dropped slightly further from the previous section.
  • Margin I have kept the same as in the previous section.
Next up is Solar Roof.

Solar Roof

Past Data

There is no past data for Solar Roof, because it has only started scaling two months ago.

Future Projections
  • Roofs Deployed is not an easy projection to make. Elon has said that they plan to ramp up to 1,000 / week as soon as possible, and then onward to 5,000, 10,000, and 20,000 per week. He has also mentioned that the addressable market in the US is about 1,000,000 to 2,000,000 roofs per year, and 100,000,000 per year worldwide, which might be a bit optimistic, but we'll dissect that number a bit soon. What I've projected here I think is way more conservative than Tesla and Elon's projections, but there's simply much more risk in an entirely new product. Soon, once we get more information around the initial solar roof ramp, we'll be able to make projections with more certainty.
  • $ / Roof. During the conference call in October it was mentioned that the average price of a Solar Roof will be $35,000. To be honest, I feel like this is likely to be a lowball number to get potential customers excited, but for the sake of further conservatism in a completely new line of business, I've gone with $32.500. I have this dropping by about 1/3 by 2030, because presumably there will be further cost reductions in this new technology. Only part of the price will benefit from reductions in price of photovoltaics however, because a large part of the cost of Solar Roof will come from the roof itself.
  • Margin. 20% is what I've gone with, although it might take them a bit of time before they'll get there. I don't expect it to be 20% right now as they've only just started installing V3. Just like Roofs Desployed it's a bit of a guess though, and we should get more clarity in the coming year or two.

Potential
  • Market in Roofs. Elon said that the worldwide addressable market is 100,000,000 roofs, but I have been unable to confirm this. The best information I've been able to find on Google, is this article that claims we're going to need to build 2 billion new homes in the next 80 years. The market for roofs is of course larger than the amount of houses built each year, but I've gone the conservative route and went with 25,000,000 roofs per year. Some back of the napkin math also confirms this is more reasonable, because 1.5 million roofs per year in the USA means about 1 roof per 200 people in the USA per year. One roof per 200 people per year worldwide would dictate a market of 40 million roofs per year, which is not too far off of 25 million.
  • Tesla Market Share % I've gone with 15%, but it's a pretty personal thing. I can't blame you for going with a lower or a larger number. 5% market share would be a little over 20k roofs per week, which is something Elon mentioned is supposedly in the pipeline.
  • $ / Roof I have it dropping slightly further, but I could see this number being too high and having to come down for the technology to scale globally.

Total Energy Potential

Before we move on to Tesla's next business, I want to give you a quick impression of what kind of valuation and stock price this potential could lead to for just Tesla's energy division. If you look at the "Total Potential" tab, you'll see that I've estimated OPEX (Operating Expenses) to be 11% of total revenues. I think this strikes a good balance between conservatism and optimism, because although Tesla's OPEX is currently sitting at 14% after their tremendous Q3'19, 1/3 of this is R&D expenses which should come down significantly over time as a % of revenues. Tesla's SG&A is the main OPEX expense and it currently already sits at below 10%, and is very likely to drop further over time as Tesla grows. One therefore might be of the opinion that OPEX at 11% of total revenues is too conservative, but even at 11% Tesla would already achieve pretty amazing operating margin in this model, so I've decided to keep it at that.

As valuation method I've chosen an EBIT multiple, and I've gone with 20x EBIT. When looking at potentials, a multiple of around 10 might be more appropriate, but most enormous (tech) companies sit at an EBIT multiple of 20-30, so I think 20x will give us a better indication of where valuation will end up. For shares outstanding I've chosen 250 million, which would be about 30% dilution from today's outstanding shares. For comparison, Amazon has diluted about 1% annually over the past few years probably due to employee stock compensation, but Tesla will dilute a lot more because of Elon's compensation scheme. Furthermore, Tesla won't achieve full potential in the 2020s, and we're more likely looking at the 2030s before a lot of these numbers could potentially be hit, so it's better to factor in some extra dilution between now and then.

The following valuation and accompanying SP (stock price) in various scenarios are calculated by looking at what % of profits come from the individual business (in this case Energy), and then taking that same % of the market cap and SP. In reality, different businesses would end up with different multiples, and different business will be accompanied by different amounts of OPEX, but again this is just an indication and to give you an idea of potentials.



Hopefully you've understood that I'm not arguing that Tesla's energy division has the potential to be a $3.4T market cap. That $3.4T market cap could in theory be the market cap of one company that has a complete monopoly of energy storage, the solar market, and a roofing market where every roof worldwide is a solar roof. The $1T market cap is what I estimate the potential of Tesla's energy division to be, but you can put your own estimations into the yellow fields and see what pops out for you.

We'll look at what valuation and stock price the model with projections from now until 2030 will lead to later, together with all of Tesla's other businesses once we've created financial models for those too.

Automotive

Now that we're getting the hang of this, we can move a little bit quicker through the steps and the model. For the automotive model I am not separating passenger vehicles from cargo vehicles. I'm also not going explicitly over how the past data was found, because it works very similar to the past data from Energy. What I will do however, is go over each part of the model in detail, because it's rather large and probably a bit overwhelming.

Vehicle Projections

This is what the model for MSX (Model S + X) looks like:

  • # stands for the number of units sold in each year.
  • Lease % stands for the % of leased vehicles.
  • Lease # is the number of leased vehicles.
  • Non-Lease # is the remaining number of vehicles sold directly.
  • ASP is the average selling price of the vehicle.
  • CoGS are the Costs of Goods Sold aka what it cost Tesla to make, deliver, and sell that vehicle.
  • Margin is margin.
  • Revenue is revenue.
  • Profit is profit.
This same MSX model is also used for Semi Truck, Cybertruck, and Roadster 2, but the M3 (Model 3) and MY (Model Y) have slightly different models:


As you can see these have the total unit numbers divided into three factories: Fremont, Giga 3, and Giga 4. This helps to project how Tesla will scale these vehicles over the next couple of years based on the factories' projected outputs.

Sales Figures

This part of the model sums up all the sales numbers, which include regulatory credits. A simpler model could be made without accounting for reg credits sales, but over time I've worked to include these, because it leads to a more accurate model.


These should be fairly self explanatory. The way it works is that I estimate the ASPs based on comments from Tesla and past data. Tesla has always reported the amount of credits they sell (although they changed the way they report it in 2019, hence all the red fields), so one can then add the credit sales to revenue from vehicle sales, and get total automotive sales.

Leasing


For more details on how the leasing model works, please refer to this blog I wrote a few months ago. But in essence how it works is by assuming leasing revenue to be correlated to the value of their operating lease fleet ("op lease"), which can be found on Tesla's balance sheet.

Total

This is where the number from leasing and sales come together. It should be pretty self-explanatory.


Below these are also some growth numbers and the cumulative amount of vehicles sold with FSD capability, which will come in handy later in the Autonomy and AMaaS models.

Future Projections

For the next four years until 2023, I've modeled out the automotive segment based on Tesla's announced expansion plans and new model release schedule:


MSX

  • #: There's a high likelihood of a MSX Plaid mode coming out in 2020 which has the potential to boost sales, I could also see Tesla refreshing MSX around 2022 for the Model S's 10 year anniversary, but I've kept these numbers on the conservative side at 70k per year, which is what they're selling as of right now.
  • Lease % is hovering around 15% as of the most recent quarter.
  • ASP is close to $95k as of Q3'19, but for all the ASPs in this model I have cut off $1,500 starting in 2020, because I want to take out the FSD option revenue that I am modeling separately in the Autonomy model. I've chosen $1,500 because I'm guessing FSD attach rates are around 20% at the moment. This is based off of Q3'19 call comments made by Zachary, the CFO. He said there's about $500-600M in unrecognized FSD revenue, which considering there were about 500-600k FSD capable cars on the road at the time means about $1,000 of unrecognized FSD revenue per car. The package is going for $7,000 now, but was sold for $5,000 earlier this year, so assuming FSD ASP of a bit over 5k would indicate a little less than one in five bought it, or in other words around a 20% attach rate, which means about 20% of $7,000 of current ASPs are due to the FSD package, which is about $1,500.
  • CoGS for MSX are around $72,5k currently. I have these dropping slightly over time, because we can assume they'll improve production efficiencies over time.

M3

  • Fremont #: Fremont is capable of, and currently producing at a rate of 350k M3 annually.
  • Giga 3 #: They seem very close to 3k / week already as discussed before under BAMF Manufacturing. I assume they will ramp up to 5k / week later in 2020, because demand should easily be there, and they are currently building additional stamping infrastructure for M3 at Giga 3.
  • Giga 4 #: Tesla has said they will start building out MY production first in Europe, so I expect M3 production to come online a little later. Total Giga 4 capacity is rumored to be ~750k, but I expect more than half of this to be MY.
  • Lease %: I am assuming will catch up to MSX.
  • ASP: Current numbers minus the $1,500 FSD package.
  • CoGS: Current numbers that drop over the next few years. Giga 3 and Giga 4 are expected to be much more capital efficient as discussed during BAMF Manufacturing.

Little sidenote: Some of these current numbers are based off of my quarterly model. You can find that model, and some models of other TMC members, over in the TMC Near-future quarterly financial projections thread here.

MY

  • Fremont #: Elon has said 1k / week by mid 2020, and I'm assuming they'll reach 5k / week around the end of 2020, and then catch up to Fremont M3 production sometime in 2021. Nobody knows their plans for sure, but this seems reasonable to me.
  • Giga 3 #: Tesla has already started construction on the Model Y production phase of Giga 3 and is progressing nicely. They should definitely be able to start towards the end of 2020, and ramp fully in 2021.
  • Giga 4 #: Giga 4 is expected to go online in 2021 and will start with Model Y. This model assumes they will aggressively ramp up to 10k / week by the start of 2023, probably around 18 months after the factory initially started production.
  • ASP: Should over time be around $4k higher than M3.
  • CoGS: Are rumored to be similar to M3, because it's basically the same car but with a lot of improvements such as the one-piece body and revolutionary wiring system. I've erred slightly on the side of caution and assumed CoGS to still be slightly higher than M3.
Little sidenote: These M3 and MY numbers are on the aggressive side. I am quite confident that demand is there, and that Tesla is able to ramp production about as fast as this (Giga 3 so far has given me more confidence), but there is one uncertainty. That is battery supply. Tesla will need to scale their battery supply just as aggressively as their vehicle production. We should hear more about their plans for this during Battery & Powertrain Investor Day.

Semi Truck

  • #: Semi is supposed to go into production in 2020, but Tesla is rumored to use the first trucks themselves for their own logistical needs. The ramp up I'm projecting afterwards is very conservative, but it's an entirely new type of vehicle for Tesla for an entirely new type of customer. I expect them to do a lot of validation and take it slower, but I don't blame you if you're more optimistic than me here.
  • ASP: We're unsure what model mix will be, but the price of the SR version will be $150k, and price of the LR version will be $180k. I've chosen $165k ASP in between those for now.
  • Margin: I've chosen 22.5% on the conservative side, compared to the ~25% that it looks like Tesla will achieve on passenger vehicles long term.

Cybertruck

  • #: A conservative ramp up for the first two years. The body will undoubtedly require very different production techniques, so there could be complications there. The 500k in the model in 2023 is more optimistic than I actually am, but I've adjusted this upwards a bit to make the overall growth in that year more in line with what I think it will be. I don't think Cybertruck will sell this much in 2023, but perhaps a new factory for Model Y in China will, or a completely new model.
  • ASP: Long term Cybertruck ASP of $55k is calculated from Elon's tweet.

Roadster 2

  • Nodody really knows for sure when it will go into production, and what volumes and numbers will be, but this is what I've gone with. It should be a higher margin car, because it's very high-end.

Automotive Sales


This is basically just adding credits to the vehicles sales. Due to the FCA deal there should be a very large jump in credits in the next two years, but after this deal it might drop down a bit again depending on how well the rest of the industry is complying with regulations, and on how badly they need Tesla's help.

Automotive Leasing


I'll skip over this, because it's not that interesting to talk about. The numbers pretty much extrapolate numbers from the past.

Total Automotive 2023



All of this would lead to a production of 2,68 million cars in 2023. If you are skeptical about this, I'd just like to remind you that this is a 10x increase from 2018 to 2023. Tesla increased its production by almost 12x in the past five years from 2014 to 2019. As long as they can scale battery supply, I'm fairly confident they can achieve this, or at least come close to it.

Total Automotive 2023 - 2030


For this time period I don't think it's useful to project it in exact model #s. It's unclear what vehicles Tesla will launch in the latter half of the next decade, so I've simply projected off of growth numbers. 
  • #: These growth numbers are much smaller than what Tesla has achieved in the past, but it does become more difficult to grow as fast beyond a certain point. If Tesla really manages to master BAMF Manufacturing, there is a chance they could exceed these numbers, but 20 million cars sold in 2030 would already be very, very impressive.
  • Revenue / car: It's unlikely that Tesla will sell 20 million at an ASP of $60,000 ($63k ASP in 2023 includes ~$3k in leasing revenue and ~$400 in credits revenue), so I have this dropping pretty aggressively to $45,000 per car.
  • Margin I also have this dropping a bit to 22.5% to be a bit more conservative.

Automotive Potential

To figure this out is a bit trickier than just finding the number of cars sold per year, because the industry will change drastically with the advent of autonomy as I explained in my piece on the Automotive Industry's Transition to AEVs:

Currently, the situation is as follows:
  • An ICEV lasts close to 200,000 miles
  • The average miles per driver per year is 13,764 in the USA7,900 in the UK, and about 7,500 in the EU. So let's say the total average is about 10,000 miles per driver per year.
  • This means that the average vehicle will be replaced every 20 years. This matches up with comments from Elon that it takes 20 years to replace the global fleet of 2B vehicles.
This means that as of right now, the total automotive industry sells 2B vehicle per 20 years, or an average of about 100M vehicles globally per year.

In the AEV future, the situation will be very different:
  • EVs have less moving parts, and therefore last longer. Tesla is working on a 1,000,000 mile battery and powertrain.
  • Robotaxis could drive up to almost 100,000 miles per year. This means they would last approximately 10 years.
  • Autonomy will do away with private car ownership, because autonomy will massively decrease the cost of MaaS, and at the same time massively increase the up front costs of vehicle ownership.
  • The advent of autonomy will drastically change the form of our vehicles. Private car ownership means cars have to be able to handle a wide variety of use cases. Widespread MaaS will lead to more optimized vehicles, such as small 1-or-2 seat vehicles with limited cargo space for a large portion of taxi rides. I expect that as a result the average cost of vehicles will drop significantly compared to what it is today.
  • The high utilization rate of an autonomous robotaxi will mean we need far less vehicles on the road. I've seen figures that said we'll need 5 times less vehicles, but then again cars are parked 95% of the time, so maybe we'll need even less than that.
  • The drop in cost of transportation that AEVs will bring with them, will lead to a larger demand for transportation, as long as our road network can keep up. I expect AEVs will mainly transport passengers during the day and peak hours, and mainly transport cargo during the night and off-peak hours. This is also why we need TBC (The Boring Company).
Therefore, we will calculate the market size of the automotive industry through VMT (Vehicle Miles Traveled) worldwide. The 2 billion total vehicles on the road today from above appears to maybe be too high, because this article says we're around 1.2B today and will grow to 2B by 2035. Sticking with the 10,000 miles driven per vehicle per year, we'd get annual VMT of 20 trillion. But an autonomous shared vehicle will spend time driving from one job to the next, and the useful miles of an AV are likely only about 50% (similar to Uber drivers). That'd increase global annual VMT to something closer to 40 trillion miles, and assuming there will be some increase due to the price of transportation dropping drastically, I've gone with 50 trillion annual global VMT in an autonomous shared mobility future. This would not be by 2030 though, but more likely 2040 or 2050.


With a vehicle longevity of the promised million miles this would lead to annual vehicle production of 50M vehicles, which is about half of what it is today. This could drop further if vehicle longevity improves beyond 1M miles, but it could increase further if the reduction in cost of mobility turns out to increase demand even more.

I've put Tesla's potential market share at 60% because of all the things discussed in the Automotive section of this blog, but you can put in what you think will happen.

All of this gives Tesla's automotive business the following potential market cap and SP given previously discussed valuation metrics.



As mentioned, discussion of the 2030 numbers will be done at the end after we get through Autonomy and AMaaS.

Autonomy

I did not have an autonomy model when I started writing this blog, and just finished creating it. I might add additional details in the future, especially if Tesla starts reporting more information about FSD attach rates, but for now this is what I've come up with. There is no past data, because Tesla has never reported any detailed figures, it's purely a model to project the future and to help us understand what Tesla's autonomy business could do in the future.

Future Projections

  • Vehicles Sold is the number of vehicles sold in said year, and the number is taken from our automotive model.
  • FSD Attach Rate is the % of people that add the FSD option. As discussed previously it appears to be around 20% right now, but as the software improves over time I expect more and more people to buy, especially because they've mentioned they expect the price to rise steadily as the software becomes more valuable. This model assume that by 2024 FSD will be as safe as a human and will have regulatory approval. At that point I'd expect attach rates to be 100%, probably in a similar way to Autopilot that has now become standard on every car. Autonomy could be solved sooner, it could be solved later. Either way it doesn't really influence the 2030 numbers, but you can adjust it if you want to.
  • FSDs Sold is simply calculated from the two numbers above.
  • FSD Price has gone up from $5k to $7k this year alone. I've assumed similar price increases over the next few years, and a bigger bump in 2024 when it gains regulatory approval. If you think $20k for a software option is expensive, I suggest you consider how much an AV can earn for the owner. What autonomy is is basically a 24 hour Uber driver that works its ass off for you, that never has to sleep, and only has to take a break when the vehicle needs to recharge. It could easily be sold for more than $20k, because this software can make tens of thousands of dollars each year driving passengers and cargo around for the customer.
  • Revenue is simply calculated from the two numbers above.
  • Costs and Margins. When you sell software, the costs are basically the same no matter whether you sell 1 copy or whether you sell 100 million copies. Because of this software margins tend to be very high, and can be as high as 95%. Adobe has 88% margins on their software for example. For Tesla's FSD software I've gone ahead and made some cost assumptions that are probably too conservative, but margins still end up being 95%. I've assumed that about 1/10 of Tesla's current R&D budget is spent on FSD development, half on software and half on computing infrastructure*. Then I've assumed that Tesla increases Autopilot engineers by 20% each year, and that computing infrastructure stays the same as a % of revenue. You can see that even with these probably too high cost projections, margins end up being 95%. The only thing I've been able to think of that would add more costs and lower margins, is if Tesla decided to bundle some sort of insurance into the software. If not for the fact that Tesla as a growth company really values the up front cash, I could even see them selling FSD as a subscription model. Anyway, it's hard to predict these with certainty, but you can adjust margins if you think 95% is absurd. It's a mass market piece of software costing $20k though, so to me it doesn't seem far fetched. And before I forget, there is of course a cost related to the hardware (sensors and computer chip), but these are already accounted for in our model in the automotive section.
*These costs would probably end up under OPEX actually, but I didn't want to just flat out put down 100% margins.

Autonomy Potential


You don't have to fill in your projection in this yellow field, because it automatically copies it from the one you filled in in Automotive Potential. After all, in the distant future all vehicles should in all likelihood be autonomous, especially when comparing a human driver to an FSD system becomes the equivalent of comparing a drunk driver to a sober driver today.

I really think the $20k price for an FSD system even if there is competition is very reasonable. It's just an extremely valuable piece of software. I think it's too complex a product for it ever to be commodotized, and Tesla's current trajectory will likely keep them ahead in terms of safety even if others do also solve autonomy.

I've kept margin the same. Costs should decrease if anything, but 95% is already really high.

All of this leads to the following market cap and SP potential for Tesla's autonomy business:



In case it's not obvious, these valuations and SPs are for individual business, not all business combined. We'll get to that soon. First we'll model out AMaaS.

AMaaS

This model I have improved tremendously since my original Investment Thesis from six months ago. It's gotten quite complex and detailed, so I'm going to explain this one step by step.

Total AVs


  • Total AVs EoY is the total cumulative number of FSD capable vehicles Tesla has produced at the end of that year. This number is taken from the very bottom of the Automotive model.
  • Avg AVs in year is calculated by taking the middle between the EoY number from the previous year and the EoY number from this.

Tesla Fleet

Tesla has said they will operate their own fleet of vehicles in addition to allowing customer vehicles on the Tesla Network.

  • Profit in previous year is taken from the "Total" tab which we'll get to later. It's the profit Tesla made in the year prior.
  • % of Profit into fleet is a guesstimate of what % of its previous year's profits Tesla will put into buying AVs for its own fleet.
  • Profit into fleet is calculated from from the above two numbers.
  • ASP fleet AV is an estimation of the ASP of the vehicles that Tesla will add to its fleet.
  • AVs produced in year is taken from the Automotive model. It's the umber of vehicles Tesla produced that year.
  • AVs into fleet is calculated from the above three numbers. It's the number of vehicles added to Tesla's fleet that year.
  • AVs in fleet end year is self-explanatory.
  • Avg AVs in fleet in year is once again an average. The middle between AVs in fleet end year from last year and from this year.

Customer AVs

  • Avg cust AVs in year is calculated by subtracting avg AVs in existence in that year by the number in Tesla's fleet.
  • Participate in network % is the % that will participate in Tesla Network.
  • Cust AVs on network is the actual number calculated from the two values above.

Network Stats


A lot of these are self-explanatory, but these might not be:
  • Total AVs on network is combined Tesla fleet and customer AVs.
  • Useful Miles is the % of miles that can be billed to customers. Some miles will be driven from one job to another.

Network Financials


Most of these numbers are calculated from the numbers above. Only "Tesla Cut" is an estimation of the cut Tesla will take from customer AVs on the network.

At the very bottom you can find the "Total AMaaS Revenue", "Total AMaaS Profit", and "Total AMaaS Margin".

With that out of the way, let's get into the numbers.

Future Projections

  • % of Profit into fleet I have ramping up quickly to 75%. As you'll see soon, there's a massive ROI on owning a fleet of AVs (hence $20k for FSD system not being expensive), so it makes sense for Tesla to 'buy' as many of their own vehicles as they can.
  • ASP Fleet AV I start off at $60k, and have it dropping to $50k by 2030. I'd expect Tesla to launch cheaper models in the late 2020s, although they might not because the economics of an M3 or MY as a robotaxi are already fantastic. Important: This ASP includes $20k for the FSD package.
  • Participate in network % I also start off conservatively. I've explained previously how I could see them ramping up AMaaS a bit slower.
  • Avg Hrs / week means 14 hours per day, which is less than the 16 Elon estimated during Autonomy Investor Day, and I expect that one could reach far higher utilization by delivering cargo during non-peak hours.
  • Avg MPH is conservative and much lower than what Uber reports today at 25-27 MPH.
  • Useful Miles is kept the same as we talked about before, although some articles claim Uber drivers see better percentages than this.
  • Cost / Mile & Revenue / Mile: Elon claimed in his Autonomy Investor Day presentation that the cost of operating a robotaxi would be $0.18 as of 2019, and that they could charge $1 per mile, which would significantly undercut Uber/Lyft's average price per mile of $2. However, those numbers lead to absurd revenues and profits, so I've made this model far more conservative starting at $0.15 and $0.5, and dropping to $0.1 and $0.33 by 2030. Margins this way start at 24% and end up at 33%, which is still high, but more reasonable. It's very possible that Tesla can charge more per mile if they want to, because they might be the only player, and even if they're not they'll have massive cost advantages. This is what I went with, but as with everything you can adjust them if you'd like.
  • Tesla Cut I've put at 50% of profits, which would only lead to an about 20% cut of revenues. This is far below the 30% of revenues Elon talked about during Autonomy Investor Day, and also less than Uber takes. Uber takes 20% of revenues at the very minimum, but more often than not takes 25-30%, and in some cases as much as 50% of revenues. If you want to be more conservative for some reason, you can put 25% here, which would lead to only 10% of revenues.

AMaaS Potential


Once again a fairly complicated model, but I'm sure that at this point you can figure out by yourself how it works, so I'm only going over the assumptions/guesstimations:
  • Market in Million VMT / Year is taken from the Automotive Potential model.
  • Market Share potential I put a little lower than the automotive market share to account for the possibility that there will still be some VMT that are not through a ride sharing network. Especially in rural areas private vehicle ownership might still make some sense. Again, you can fill in a different number to see what'd happen in that scenario.
  • Useful Miles % I increased slightly, because a large network should be able to find efficiencies in this aspect.
  • Price / Mile could drop a bit further in the very long term, so I reduced it by about 25% from the other AMaaS model.
  • Cost / Mile I kept the same as the other AMaaS model, just to be a bit conservative.
  • Tesla Fleet Contribution to Network at 75% seems conservative to me. In the long term Tesla (according to this model at least) should be printing cash and buying a very large percentage, perhaps all of its own vehicles. In 2030 in the AMaaS model this is already at about 75%.
  • Tesla Cut I've kept the same, but you can reduce it if you think there will be a lot of competition. Even with 50% of the profit going to Tesla, customers should still be printing money on the Tesla Network though.
This leads to the following market cap and SP potential.



Once again, these are for individual business, not cumulative.

We'll look at Tesla's total potential very soon, but first let's look at the result of the 2020 to 2030 financial model we've created.

Tesla 2020 -> 2030


The assumptions/estimations:
  • OPEX: There's probably a few people who make their financials models who'll think these are far too high. Tesla showed extreme improvements in operational efficiency in Q3'19. This model assumes more rapid OPEX expansion than many other models I've seen, but I've kept it as a % of revenue after 2022 going down to 11% from today's 14%. Considering R&D should drop as a % of revenues significantly over the next ten years, and SG&A is already under 10%, 11% is VERY conservative in my opinion. One small note is that I've excluded Autonomy revenue from OPEX calculations, because whether Tesla installs FSD software on 1 of its cars or on 10 million, should not bring forth any changes in OPEX.
  • Valuation: I've decided on an EBIT multiple for 2020 of 30x that drops down to 20x by 2030. Again, I believe this is conservative. Most large tech companies trade at significantly higher multiples (Amazon is currently around 80x, Alphabet 30x, Facebook 35x, and Apple 25x).
  • Shares Outstanding I've increased by 4 million shares per year for the next two years, because currently 1 million shares per quarter are added. After that I've accounted for 2% dilution per year to account for the CEO Compensation Package that should be paid out sometime in the next decade.
The result of all this would be a SP of $58,246 in 2030.

Tesla's Potential


According to my estimations, Tesla has the potential to achieve a $28T market cap, and a $112,886 SP.

Now, I'm most curious about what kind of market cap and SP your estimations lead to, so if you've followed along and filled in your estimations in the spreadsheet, please leave a comment with your valuation and SP.

Financials Conclusion

Rather than argue here about exactly what Tesla should be worth today, next year, in 10 years, or in 20 years, I'd rather focus here on the overall lessons we can learn from these models. Financial models that go 10 years into the future, or even further than that and attempt to project total potential, are not meant to be precise predictions of how the future will unfold. Rather, they are meant to learn more about how a company works, its business, and the keys to its success.

Therefore, if you want to learn more about Tesla, I highly recommend downloading this model and playing around with it to see how various things impact the company. There are dozens of lessons that can be learned, but here are just a few of my personal key takeaways:
  1. Tesla's automotive business is as fundamental to its trifecta of Automotive - Autonomy - AMaaS as its batteries are to its overall business. If you play around with the Automotive model, and put in much more bearish/bullish predictions, you'll see that it heavily impacts the autonomy and AMaaS businesses. This is because every FSD capable vehicle produced can, and eventually will be upgraded by the autonomy business with a FSD package, and because every vehicle sold can contribute to the Tesla Network.
  2. Tesla Energy may not have the potential to be as large as any of its other businesses, but if somehow Tesla screws up royally in those, the battery storage and solar roof business still can still be a saving grace. One trillion $ in market cap may not be large compared to a total market cap of $28T, but it's still enormous. Especially battery storage which Tesla is in an amazing position to crush because of their battery lead, could be worth $300B in revenue and $60B in profits according to my potential estimations.
  3. I've completely underestimated Tesla Autonomy's potential until now. Like I mentioned, I created the Autonomy model while writing this blog, and I never realised before how big this could be. I thought it'd just be great because of the implications for robotaxi/AMaaS, but Autonomy is actually a very serious business on its own, and I think the $20k for a regulatory approved, as safe as a human FSD package is potentially a lowball number too.
  4. Q3'19 really was Tesla's best quarter ever (so far...). Six months ago I had similar OPEX numbers, but back then they were optimistic. After the operational efficiency improvements Tesla showed in Q3'19, an OPEX projection of 11% of revenues suddenly looks pessimistic. Watching how OPEX progresses is one of the things I'm most excited about in Tesla's near term financials.
  5. Tesla's automotive business is not an ordinary automotive business. Of course Tesla has the other three businesses on top of its automotive business, but even its EV business alone is completely unlike the rest of the industry's ICE businesses. The margins on Tesla's vehicles and its operating margin are superior to that of the mass market ICE manufacturers of today.
  6. The future Automotive - Autonomy - AMaaS market is enormous. Worth about $50T in total. There are some assumptions hidden within this, such as all future road transportation being of the AMaaS kind, but even if you cut this in half or reduce it by 80%, you still end up with a $25T or $10T total opportunity.
  7. The five most important things to Tesla's success are maintaining its lead in batteries, Elon continuing to lead the company, a large scale solution to its battery supply chain, solving autonomy, and BAMF manufacturing. This is why I'd love to see a Manufacturing Investor Day in late 2020 or early 2021 after the Battery & Powertrain Investor Day. Please make it happen, Elon!

Final Thoughts

It's now late afternoon on Monday the 30th of December 2019, and I just spent about eight days non-stop writing (12+ hrs / day). I'll probably need another day or so to go over everything to correct errors and polish things, but what a crazy project this has become. I think that including some hours brainstorming before I started writing, the entire project has easily cost me 100+ hours.


Pasted in a word document it ticks in at more than 130 pages, and according to wordcounter.net I've written a total of nearly 34k words and nearly 200k characters. Woow.

Anyway, I hope you've enjoyed reading this, and I hope you've learned some things. I did not expect I'd be explaining how to solve autonomy, nor that I would give a flash course in financial modeling. So now you can go out and get a job as a self-driving car engineer or as a Wall Street analyst, all thanks to me.

Alright, time for me to get some rest, and time for you to drink some water, because you've probably been reading for the better part of a day if you haven't taken any breaks.

Happy New Year in advance! It's shaping up to be an amazing 2020 for Tesla, and I hope your 2020 will be just as grand!

Until next time,
Frank

P.S. Please share this with friends, family, and with any communities you think might enjoy reading this. I'm very passionate about Tesla, and I'd love for as many people as possible to learn more about Tesla.

Tesla Investment Thesis 2.0 Financial Model - Excel
Tesla Investment Thesis 2.0 Financial Model - Numbers
Tesla Investment Thesis 2.0 Financial Model - PDF

Comments

  1. This is David Zhao from TMC! I am so sorry that I didn't get the chance to reply on TMC last time but I have taken your advice about options seriously!

    What a great blog article again! I also had the same "Monopoly" thought in Dec 2019 and was wondering if it is going to be like "Google vs nobody" or "iPhone vs Android". I also thought after FSD is matured Tesla might have developed or enabled the core technologies for consumer grade robots, which will be another era.

    I will read the blog a few times and leave more comments later! Happy holidays!

    ReplyDelete
  2. Would be interesting if you played the devil's advocate and come up with a short thesis on Tesla also. Great post, thank you.

    ReplyDelete
    Replies
    1. I'll consider this. Thanks Geary!

      Delete
    2. Geary,

      I'll still consider this for a future post, but I just stumbled upon this:

      https://www.dropbox.com/s/rmbrlmq6vxzchc8/Stanphyl%20Capital%20Letter%20-%20December%202019.pdf?dl=0

      It's a letter from Mark Spiegel to his investors, and from pages 7 to 16 he has written his thesis on why Tesla is a good short.

      Delete
  3. Finished a pass and probably will read again soon. Here are some thoughts:

    1. Software, softwarization, and virtualization.

    I actually think this is a core advantage of Tesla. Putting a layer of software virtualization on top of the hardware and sensors in the car allows fast improvement and iteration, which is common in SaaS software and uncommon in cars.

    The same is true for "Machines make machines" (factory). I think they have virtualized (softwarized) or they are virtualizing the factory too.

    A recent example is the Cloud Computing, which only became possible after major virtualization development of physical computer servers. (VMware)

    I also think the software and programming nature of the silicon valley community had a huge influence here since I work here.

    2. My 2020 goal is to find a job in Tesla as a Software Engineer. :)

    "And if you are a talented, motivated, and hard working individual alive today, there is no better place to work than Tesla"

    And I heard from many people online or in person that without stock appreciation, Tesla's pay for a regular engineer is about 20% lower than Google and Facebook in the bay area (silicon valley). And the work life balance is far worse than Google, on par with Facebook maybe.

    3. Tesla's long term potential could be affected by the raise of VR.

    VR may decrease the demand of transportation in general. (People dont need to travel or commute to work)

    ReplyDelete
    Replies
    1. 1. I agree, software is becoming more and more important. Tesla indeed has a huge lead here compared to traditional automotive companies.

      2. Good luck! You'll indeed have to accept a terrible work life balance, but there are a lot of upsides to it.

      3. Possibly? But then you're arguing that we'll all just sit at home and lead virtual lives? I don't think that'll go that quick in the near term, but Elon does talk a lot about living simulations.

      Delete
    2. For VR, here is my estimate

      1. By 2021, high-resolution consumer VR headset is open for sale
      (current the resolution is the most restricting factor for VR to be more useful)
      (Facebook Oculus is the obvious leader and if Sony and Microsoft can catch up, they will lose the game market.)

      2. By 2023, VR headset becomes popular like phones.
      We might start to see Xiaomi, Apple in the VR market, other than Facebook.

      3. By 2025, VR starts to affect the daily life and consumer pattern like what smartphone did. This is the time that it may affect Tesla, or the transportation pattern that people have, in general.

      I have a few VR sets and I follow the trend closely. I have fair confidence in the estimate. :)

      Delete
    3. I don't really follow VR, so I don't know much about it. But I'm skeptical that it will play out exactly like that:

      1. There are people (including me) who feel unwell after using a VR headset for an extended period of time. I once did a demo for just a few minutes, and needed about ~5 minutes to recover after, because I felt so nauseous.

      2. I don't think high-resolution is enough for people to want to spend the majority of their lives in VR. It'd have to be indistinguishable from reality, which I don't think is going that happen that fast.

      3. People need people. People need person-to-person social interaction. People won't just sit at home doing everything through VR, because they need to go out to spend time with other people in order to be happy.

      Delete

    4. All are fair points!

      But if you trust me (I follow your blog closely and i am a long term Tesla investor)

      Just go ahead and buy this, you can’t miss it, it is another Tesla level innovation.

      Oculus Quest All-in-one VR Gaming Headset - 128GB https://www.amazon.com/dp/B07PRDGYTW/ref=cm_sw_r_cp_api_i_3GFhEbM8TMRVZ

      Delete
    5. The VR link is from me, I am on mobile, David ZHAO

      Delete
  4. Another 2 comments

    1. Have we considered other factories other than the Shanghai and Berlin ones?

    2. It would nice if any popular Youtubers can make some videos out of any topics here. HyperChange, Now you know and so on.

    3. Have we considered the implications of the 2018 CEO Reward? There are 12 milestones and I thought it might take 1 year for each milestone on the ballpark.

    https://ir.tesla.com/static-files/55362f0a-ee8a-4fcc-ba11-cc09194974b6

    ReplyDelete
    Replies
    1. 1. Tesla will undoubtedly announce more factories in the future, or else they'll have to stop growing. I think they've briefly mentioned having multiple factories in China long term, as well as a factory in the east coast of the US.

      2. Yeah, that would be cool! I tweeted this blog to Galileo :)

      3. Yeah, I've tried to account for this by modeling out a fairly high amount of dilution.

      Delete
    2. Super cool, I am looking forward to seeing Youtubers start to talk about this blog!

      Delete
  5. There's one thing missing here, the APP-Store - how is Tesla not already having third-party application freemium models in an app store and taking revenue-share like apple? i.e. each country has it's own legislation for company-car-usage-tax-regain (sorry, non native speaker) - so let the already existing companies for comany-car-usage software and hardware get access to TEslas hardware in the care and charge for the service ... and there's sooo much more usecases for apps specific for the car ... delivery services etc...

    ReplyDelete
    Replies
    1. oh - i forgot the abetterrouteplanner app that I'd looooveee to pay a few bucks every month for in my Model S !

      Delete
    2. Yeah, there's a lot of potential around infotainment, including them creating an app store. Looking at the things they're doing with gaming such as Stardew Valley, I wouldn't be surprised if Tesla is already thinking about and working on this.

      Delete
  6. I disagree with the statement that batteries will not be viable for long haul shipping. Container ships doing trans-pacific routes only need about 2.5% of their payload to be fuel at 20 knots. That fuel has an energy density of about 11,500wh/kg. About 50x batteries. But the fuel is only burned at 40% efficiency. An electric motor purpose made will get 96% so 40/96 * 50x gets you to a battery that’s about 20x the mass of the heavy fuel. This means your payload is cut in half if you are mass constrained, the effect will be less if volume constrained.

    50% is a significant hit to payload but this does not mean it’s infeasible, and this is a worst case trans-pacific route. New York to London would be more like a 15% penalty. Totally doable, the battery will be a bit pricey right now but these ships cost over 1 billion to start.

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. Thanks for your insights!

      I admit I did not do any calculations myself. That would've taken a lot of extra time that I rather spent on other sections, so I just researched the possibilities through Google and various articles. I only stumbled upon articles that showed a few calculations and said it was impossible.

      If that 2.5% is true, it sounds like you could be right. I also think your point about motor efficiency is fantastic, and I had not heard about that one before as to how it relates to the electrification of shipping.

      Delete
  7. Now long haul aircraft... that’s not doable. They are ~40% fuel by mass at takeoff.

    ReplyDelete
    Replies
    1. Elon has already made designs, and says that it is possible at 400-500 Wh/kg. Perhaps these are somewhat shorter distance aircraft, but unless proven otherwise, I will trust his numbers on this.

      Delete
  8. FSD is ANI.

    "Alien Dreadnought" will be built for Cybertruck, and the first generation of robots with ANI will be used in it. It can drastically reduce the number of workers needed in the factory, which means reduce the building cost to very, very low level.

    If robots with ANI can build trucks, then not many things they cannot build. For example, automatic natural resource mining, automatic tunnel digging, etc.

    How much do these robots with ANI worth?

    Tesla cars and trucks are not just products, they are the means to reach ANI.

    ReplyDelete
  9. Frank

    Phenomenal work. So much diligence. So much information behind this I still have to read too.

    I am going to sound like a bear. I have invested a significant sum in Tesla and would love for your forecasts to proven correct. And let me say I believe Tesla could be a $3T to $6T market cap stock. But here is my big disagreement with you.

    I am not going to critique any of your models or assumptions. I have not done the work to earn disagreeing with you. But even if I go with your LOW case, you are expecting 1 company to have a market cap ($15T) of about 11.5% of expected global GDP by 2030 (~$130T @ historical 3.6% avg annual growth rate from now to 2030). Even if GDP growth surprises on the upside (and I think it will due to the transportation AND energy models changing and becoming much much more efficient) and the global economy is at $200T by 2030, your low estimate is still figuring on 1 company's market cap to be almost 8% of GDP.

    I think it's because you base total AMaaS market to be valued at $50T. That would be 40% of global GDP (at $130T). I don't think all transportation (road, rail, air and water with all manufacturing and energy included) today is even valued at $50T and AMaaS should make transportation cheaper.

    Did you think about this, and do you think it's an invalid point? Or do you think global GDP will be even higher than $200T by 2030?

    Thanks.

    ReplyDelete
    Replies
    1. Habakak,

      Thank you so much for those kind words! I'm glad you enjoyed this piece.

      To address your concerns:

      1) You're comparing apples to oranges. Market cap should not be compared to GDP, because market cap is the value of the company (usually based on earnings). It would make more sense to compare some of the potential revenue numbers to future GDP projections. When you do this, a $12T total revenue opportunity in all these markets combined seems quite reasonable, because it is only ~10% of your projected global GDP in 2030. I believe that today transportation and energy make up ~10% of global GDP each.
      2) The market cap numbers are high, perhaps too high, but market cap valuations usually are usually based off of earnings. If company A is valued at $1T based on $100B in profit from $500B in revenue, and then company B disrupts it, takes over the market, and profits $200B off of $400B in revenue, it could very well end up with a $2T market cap, even though the market shrank in terms of revenue. So considering Tesla looks like it's going to be an extremely efficient and profitable company in the future, certainly more so than the current companies in the transportation sector (Volkswagen, Ford, Uber, etc.), it could end up with a market cap higher than people are expecting based on comparisons.
      3) There's been some confusion around what I meant with the "potential" calculations, so maybe I did a bad job conveying that, but Tesla's total potential pretty much means a best case scenario. Furthermore, this potential would definitely not be achieved by 2030. If Tesla ever gets to a $20T or $25T market cap, we would be well into the 2030s, perhaps even 2040s, before that happens.

      Last but not least, keep in mind the little disclaimer I gave. I don't think one can accurately model out a business 10 years or even more into the future. Many unforeseen things can happen between now and then. These models are an amazing help to me (and hopefully others) to help understand the company, how different things will affect different outcomes, and to gather an idea of how things could go. However, I think it's difficult to predict with a lot of accuracy how things will play out that far into the future.

      With that being said, with the learnings from these models, I am now of the opinion that if Tesla gets stuck at a $2T market cap or less, that would be quite a disappointment considering their large potential. I'm hoping they'll get to $4-5T at least, potentially $10T+, but I'm not confident enough in being able to predict that far out into the future to say it is likely to happen.

      Frank

      Delete
  10. Frank

    Thanks for your response. Again, you are doing excellent work and the effort is amazing. I understand that making projections is hard to get accurate and the more time passes, the more things can happen to affect it.

    As for your potential calculations, I think you made it pretty clear. It is the monopoly case. Either way, I didn't even look much at that because I think a $5T market cap would already be more than crazy good for me.

    I guess the central point is the company becomes more efficient and earns more profit on the same or even less revenue. EV and RE accomplishes this very easily and that is what Tesla is about inside of a software paradigm while controlling the whole stack instead of outsourcing.

    At 50% annual growth (Tesla has been maintaining a higher rate for more than a decade and Elon hinted at as growth rate for the next decade for the company on the ER call the other night - we know it will be harder as the company becomes much bigger, but they have so many factors that can make this possible - alien dreadnought, AMaaS, autonomy, etc.) Tesla can easily reach a $3T market cap by 2030 and that might seem conservative in retrospect.

    Looking forward to your next blog entry.

    All the best.

    ReplyDelete
    Replies
    1. Thank you, Habakak! Couldn't have said it better myself.

      All the best to you too.

      Frank

      Delete
  11. Wow Frank I am really amazed and shocked at the level of detail and heart you put into your research and thesis! This is truly admirable stuff. On top of that I truly appreciate you explaining different factors and considerations you have, even as to outlying what may go wrong as an alternative perspective. Will definitely bookmark this blog as a go-to whenever I need information.

    I myself recently came across Tesla and have become an investor after spending almost hundreds of hours on research. I am blown away by their world-class manufacturing, execution and plans and have no doubt they have a huge future ahead of them. Never expected so much gems from this one blogpost! Looking forward to more from you! Although I would suggest smaller, bite-sized posts that are easier to digest in future ^^

    I wish you all the best in your future studies and I hope you manage to get into Singapore through your MBA :D

    ReplyDelete
    Replies
    1. Hey Marc,

      Looks liked I missed this comment somehow, sorry about that.

      Thank you so much for those kind words! I'm glad to hear you like my posts and find them useful.

      I've just arrived in Singapore, and will move into my apartment in the next week or so. My MBA next year will be extremely busy, but I hope to be able to make a few small posts here and there. I am very curious about the institutional investor data that will be released over the next couple of days, so I will likely make another TSLA Holders blog post soon.

      I totally agree with you on Tesla's world-class manufacturing, engineering, etc. The company has a very bright future ahead!

      Frank

      Delete

Post a Comment