July 12th, 2019 by Chanan Bos
Elon Musk shook things up on Twitter a bit this week. Some fans got concerned Tesla would stop selling cars at some point, but the core point was simply that the value of Tesla cars will go up a lot once Full Self Driving is implemented. Elon clarified that what he meant was simply that the price was going to go up significantly.
To be clear, consumers will still be able to buy a Tesla, but the clearing price will rise significantly, as a fully autonomous car that can function as a robotaxi is several times more valuable than a non-autonomous car
— Elon Musk (@elonmusk) July 8, 2019
In response, I was first going to share that I felt like Tesla backstabbed the community with this news, since this would effectively mean that a big part of the Tesla community, including myself, who cannot yet afford the car will never experience the joy of one day picking up their very own Tesla at store/service center someday. Then I started crunching the numbers and suddenly sadness was replaced by excitement. If you, too, have been feeling a bit down by this news, maybe this will cheer you up. It all started with Elon’s response to this tweet:
— Elon Musk (@elonmusk) July 8, 2019
Elon’s reaction to this tweet suggests that this course of action anticipates full autonomy before having enough factory capacity to build as many cars as the world needs. That leads to some very important questions. How many cars could Tesla end up making? How many autonomous vehicles does the world need? It is for sure quite a deep rabbit hole to explore.
How Many Robotaxis Does the World Ultimately Need?
This is the absolute most important question and it seems either no one in the media has asked this question, or just hasn’t written about it since they can’t find an answer. We don’t know the answer either, but we can present data and make some guesses.
By the time Tesla’s autonomous vehicles come out, it seems the world will have just reached 8 billion people. According to one calculation, there were 1.3 billion vehicles in 2016, and the number of passenger vehicles (excluding large trucks and buses) was just a bit under a billion. So, we know that we will need less than a billion cars.
According to Statista, in 2018 almost 79 million cars were sold worldwide. So, basically, the whole market turned into autonomous vehicles, we could replace all previous cars in 11 years. One difference, however, is that Tesla’s autonomous vehicles cannot be treated like regular vehicles — they will have several advantages. Let’s go through them.
These Tesla robotaxis are supposed to have a “million-mile drivetrain.” Even if that is an overstatement, the average car never make it to 200,000 miles — although, with extremely routine servicing, some taxis have been known to make it to 400,000 miles. In the worst case, Tesla doubles the upper limit, or it could quadruple the average.
Next, because autonomous robotaxis are shared by a lot of people, theoretically, you need fewer of them than regular cars today. The question of how many autonomous cars will be needed also very much depends on issues like: rush hour; ownership split; first/last few miles from public transport; as well as availability, privacy, efficiency, and punctuality of public transport. Do we want there to be enough autonomous cars so that every individual or family can access an autonomous vehicle during rush hour? For now, that is a question without an answer.
Assuming the US average of 832 vehicles per 1,000 people in 2016 becomes a global average for people trying to get to work in rush hour, that would mean an eventual need for 6 and a half billion autonomous cars for the whole world. Not very realistic, but let’s say that Tesla wants to build half of those cars in the next 10 years. That would mean a need for 325 million cars per year! If each Gigafactory was capable of outputting 500,000 cars per year, like Gigafactory 3, then Tesla would need 650 factories worldwide, or have each Gigafactory output 3 million cars per year and have 100 factories. Clearly, this is not a reasonable number and Elon has indicated this before.
So, let’s try this from a different perspective, a cost and capabilities perspective rather than a worldwide maximum demand perspective. This is where the latest tweets from Elon come into play. Let’s say Tesla starts producing autonomous cars that cost $50,000 to build, Tesla Model Omega, and sells them for $100,000 at a 50% profit. Then let’s use the 150 vehicles for 1,000 people global average of 2010 (not US average). It’s not a perfect number to use, but since in Elon’s vision of the future a lot fewer people use cars and people hardly own cars, it’s as good a guess as any. That would mean a need for 1.2 billion cars. Let’s say that Tesla wants to make a bit less than half of them — say, 500 million cars.
If Tesla wanted to reach 500 million cars over the course of 10 years, that would mean it would have to build 50 million cars per year. That is a lot. That is a hell of a lot. But considering how many cars get made each year, that’s a lot more reasonable. Let’s see what the numbers tell us. We need to find out:
- How much money Tesla would have to raise to fund that.
- How quickly it could raise the needed money.
- How many cars they need to sell to achieve that.
- How long it would take to build all these self-funded Gigafactories to make this happen.
To build 50 million cars a year, Tesla would need 50 GF1-type factories that can build 1,000,000 cars, or 100 GF3-type factories that can build 500,000 cars per year and each cost $5 billion to build. Alright, so that means that Tesla has to raise $500 billion, less than the annual US military budget. That is the goal, and the answer to the first question.
So, the news is that self-driving Teslas will become significantly more expensive. To reach $500 billion, Tesla would need to build and sell 10 million $100,000 autonomous cars with a 50% profit margin to fund that. Now, suddenly, these numbers start making a lot more sense — as you will see in a moment if it’s not apparent yet.
According to current trends and plans, conservatively, Tesla should be able to build 2 million cars in 2023, and according to the calculations could build a total of ~17 million between 2023 and 2026. So, let’s say that in 2023 Tesla will start building the fully autonomous Tesla Model Omega for $100,000 and will sell about 2 million of them. That would be enough for 20 GF3-type factories. That means that Tesla would have enough money to start construction of 100 GF3-type factories in just 4 to 5 years!
Now, as Elon said, you would need 100 Gigafactories to power the world. That was back in 2016 and he meant GF1-type Gigafactories, so that would mean 200 GF3-type Gigafactories. One of the biggest problems with GF1 is that, because it’s so huge, Tesla is having trouble finding enough people and the factory now is only ~30% complete. For that reason alone, it makes more sense to build smaller ones. Almost like when Elon scaled the Interplanetary Transport System down to the Big Falcon Rocket.
You may also remember that Elon mentioned the world needed 100 Gigafactories, not that Tesla would build so many by itself. Though, if Tesla is far ahead on autonomous vehicles and electric powertrains, perhaps a thought experiment based around 200 GF3-type Tesla Gigafactories is not too wild for a Friday in July.
Tesla Factory Versions 1 to 3.2
Let’s reflect on Tesla’s evolution with factories first. We have version 1 of the Fremont factory, for producing the Model S, then version 1.4 for Model S & Model X, then version 1.8 for Model 3 and 1.82 for the addition of sprung structure GA4, and then version 1.99999 for Model 3 and Model Y (if that will indeed happen).
Gigafactory 1 (GF1) for a short period of time was version 2.4, when it was announced that it would also manufacture vehicles, but as the factory seems to have had issues with expansion due to an over-ambitious size and lack of people to hire in the region, we now have design version 3, Gigafactory 3 in Shanghai.
I’m guessing Gigafactory 4 in Europe will also be a “version 3” factory since its announcement is likely around the corner. What comes next is what people are probably not yet expecting, and I think that is Gigafactory 3.2. That is Gigafactory 3 successfully ramped up with a layout somewhat better than the original GF3 & GF4 blueprint. If the theory is right, then around 2022, Tesla will announce the Tesla Model Omega, and at the end of 2023 a whole batch of new Gigafactories. Elon is probably willing to go as far as possible without getting into another “bet the company” scenario. (Those are supposed to be long gone.)
Tesla is in a hurry not because of competition but because of the nature of Tesla’s mission. It is working against the clock. While some governments are working to have the world carbon neutral by 2050, in reality, the world won’t be relatively safe if it isn’t carbon neutral by 2030.
The Machine that Builds the Machine, Built in Batches
From what I have learned from people who were overseeing large factories, having multiple identical factories is impossible unless they are built at the same time. No matter how much you may want standardized, identical equipment, companies and technologies evolve and you will want to buy or create the latest and greatest as much as possible (as much as it is financially efficient).
The same is true for Tesla, as we have seen how different the Fremont factory lines are for S & X vs general assembly 3 for Model 3, and Tesla executives have also talked (including to CleanTechnica) about improvements for Gigafactory 3 in China. As said before, the only exception seem to be for identical factories built at the exact same time. So, in other words, to achieve economies of scale and various efficiencies, we need factories built in batches, as funny as that may sound. (It is a different topic, but note that some years ago the idea of large batches of mass-manufactured satellites was laughed at, and now we have Starlink satellites sent into space 60 at a time.)
Now, let’s say completing this kind of batch of factories will take between 3 to 5 years, depending on where they are being built — let’s say 4 for the average. This is something that has never truly been done before and stands a chance of failing. Batch factory building is not exactly a standard part of university architecture and engineering courses, if you catch my drift. But let’s see what the numbers would look like.
In 2023, by selling 2 million “Omega” cars and investing 80% of that into new Gigafactories, then Tesla could earn enough to fund the first batch of 12 Gigafactories. Once that first batch comes online, the speed will really start ramping up. In any case, that means that as of 2033, you have 200 factories producing 100 million cars per year or a billion every 10 years.
Tesla Model Omega vs. Tesla Model 3
If Tesla starts making this perfect autonomous vehicle, without a steering wheel, the last model it will ever need (hence the model name Omega), in 2022 it could be pumping out 2 million cars per year. That could generate funding for 50 Gigafactories in just 2 years. Here is what a comparison of Model 3 vs. Model Omega looks like:
As you can see, while the Model 3 is profitable, without betting the company, sticking with the 2019 system would take much longer for the company to ramp up to meet global consumer needs, and Tesla would not really achieve its mission. Now I am not saying that Tesla will be able to ramp to hundreds of factories that easily just by throwing money at the problem. There are other immense challenges, like making sure the world can mine enough raw materials, manufacture enough batteries, creating a whole slew of “batch factory construction” jobs, and obtaining an adequately trained workforce. That is also why in the above calculations only 80% of the income went towards new Gigafactories.
There are a lot of ways you can change these number to even larger ones if you feel bullish or reduce the numbers by half if you are bearish. You can change the duration of the project from a billion cars every 10 years to 2 billion every 20 years, since Tesla’s aim is for its vehicles to have million-mile drivetrains. There are a lot of ways to play with the numbers. There is one really big point and that is that it does indeed make sense for Tesla to significantly raise prices once it cracks autonomy.
So, let’s not get too judgmental over the fact that we might not all get to own a Tesla. That is not a bleak future — it’s just so bright you can’t even see it yet. Nonetheless, Elon, what really comes once we have full self-driving vehicles? Where will the company go? Towards Model Omega as laid out in this article? Towards Mars? Towards electric planes? Dealing with climate change and prevention of further climate catastrophe? Right now there are more questions than answers.
As this is a really complex topic and includes some large and not very straightforward calculations, we made a very easy-to-use spreadsheet where you can play around with the numbers and see what kind of results you get. Here is the link.