Tesla's large-scale fleet learning

[diplomat33]

But Waymo and its partners would only just be starting to ramp production of Waymo's specific vehicle model or models.

Ramping to 350k units/year of custom Waymo I-Paces would probably take several years, for example.

Waymo already has a large plant in Detroit that is ready to mass produce robotaxis:
Waymo’s Detroit plant begins outfitting cars with driverless tech

So I don't think it would be that hard for Waymo for ramp up production of robotaxis.

Just scaling up the production of high-resolution, long-range, automotive-grade lidar to 100k+ units per year could take several years.

I am not aware of any evidence that scaling up production of high res, long range, automative-grade lidar would take years. The biggest problem is lack of demand. There aren't enough companies demanding high res lidar for it to be profitable yet. It's why these lidar start ups are struggling.

 

[strangecosmos2]

What's the planned production capacity of Waymo's Detroit facility? From what I can find, it's only retrofitted 30 vehicles so far. The plan is to hire between 100 and 400 employees. By comparison, Tesla's Fremont factory has 15,000 employees. I could be wrong, but my understanding is that the Detroit facility is retrofitting a tiny number of vehicles by hand. This isn't a proper car factory.

 

[diplomat33]

What's the planned production capacity of Waymo's Detroit facility? From what I can find, it's only retrofitted 30 vehicles so far. The plan is to hire between 100 and 400 employees. By comparison, Tesla's Fremont factory has 15,000 employees. I could be wrong, but my understanding is that the Detroit facility is retrofitting a tiny number of vehicles by hand. This isn't a proper car factory.

It's probably early in the ramp up, just like the Model 3 ramp up. Remember Tesla started the Model 3 with a small number built by hand before ramping up big time.

 

[strangecosmos2]

It's probably early in the ramp up, just like the Model 3 ramp up. Remember Tesla started the Model 3 with a small number built by hand before ramping up big time.

In terms of square footage, the Tesla Fremont factory is about 25x to 50x larger than Waymo's Detroit facility. The Detroit facility is 200,000 square feet, whereas the Fremont factory is between 5.3 million and 9.9 million square feet.

That said, the Detroit facility is only doing retrofits, not actual vehicle manufacturing. It's not a car factory. But this comparison helps give a sense of scale.

 

[diplomat33]

In terms of square footage, the Tesla Fremont factory is about 25x to 50x larger than Waymo's Detroit facility. The Detroit facility is 200,000 square feet, whereas the Fremont factory is between 5.3 million and 9.9 million square feet.

That said, the Detroit facility is only doing retrofits, not actual vehicle manufacturing. It's not a car factory. But this comparison helps give a sense of scale.

I don't dispute that the Tesla Fremont factory is a lot larger than the Detroit facility. But the Detroit facility will be able to retrofit a decent number of robotaxis once it reaches full capacity. All I am saying is that I don't think we should dismiss Waymo's ability to scale up their robotaxi fleet.

 

[strangecosmos2]

Some back-of-the-envelope math:

There are ~1,400 autonomous vehicle prototypes in the U.S. and, in China, I believe there are fewer. Let's say there are ~3k worldwide. Let's say the lidars and/or vehicles themselves are replaced, on average, every 2 years. That's ~1.5k units of production per year.

So, to get to 100k+ units per year, that would require a 65x+ increase in production. This is for a new technology that needs its own specialized production equipment. What's a reasonable estimate for how long it will take to scale production to this extent?

According to Luminar's timeline, it's about 3 years from the point where the lidar sensor is in test vehicles to delivering units to customers.

The difficulty is, of course, not in producing cars, since there are a lot of cars in the world. The difficulty is in a) producing the custom hardware needed for autonomous cars (e.g. lidar, battery packs, computers) and b) integrating that custom hardware into vehicles.

 

[strangecosmos2]

TBH this is really the first time that I’ve heard someone say Waymo might have a technological lead, but for some reason they (+ automotive giants like FCA) will fail because they can’t make enough cars...

This is not my argument. My argument is about which company is most likely to get to 1 million robotaxis first. I didn't say anything about “will fail”.

To produce 1M robotaxis in 3 years (i.e. 333k robotaxis per year), autonomous vehicle lidar production would need to be scaled up over 200x. Scaling up lidar production to this extent would take time.

Separately, Waymo would need to either a) spin up a mass-scale retrofitting facility or b) integrate autonomy hardware into an OEM's manufacturing process. Either option takes time.

Eligible vehicle models would likely be restricted to hybrids and all-electric vehicles due to the power consumption requirements of autonomous vehicles' computers. Battery cell and battery pack supply would therefore be a factor.

If Waymo went for all-electric robotaxis for economic reasons, that would pose an additional scaling problem.

The upshot is that if Waymo completely solves the technology problems tomorrow, producing 1 million Waymo robotaxis will likely take multiple years.

 

[strangecosmos2]

It's okay if we don't agree. You asked for my opinion:

Who do you think is actually going to win the race?

If you have a different opinion, I'm completely fine with that. Eventually, we'll all find out what happens. In the meantime, we all have the right to form our own views.

I feel I've given more than enough explanation for why I think what I think. If you're not satisfied, you're 100% free to conclude that I'm wrong.

 

[strangecosmos2]

If I may, I think a trouble spot in your communication style is that you don’t really ever label your opinions as opinions.

The words “I”, “me”, and “my” often signal an opinion. As in:

• I believe
• I think
• I feel
• I predict
• I would guess
• I am confident
• I am not certain
• To me,
• It seems to me
• My claim
• My argument
• My thought

Anytime you see these phrases, that typically signals that something is being given as an opinion. I use these phrases constantly.

You can also infer what’s intended as an opinion from context. For example, what will happen in the future (e.g. with regard to a new technology) is a topic where most of the things people say are opinions. Words like “may”, “might”, and “could” are clues that a statement is an opinion or speculation.

More examples: if a piece of writing is an opinion piece or if a forum thread is a conversation where people are discussing their opinions, then most of what’s said therein should probably be taken as opinions.

If a statement is made in a context where there are a lot of statements marked as opinions by phrases like “I think”, there’s a good chance it’s also an opinion, even if not every single sentence is explicitly marked as such. (That would be repetitive.)

Even if something is made as a statement of fact, I think (ahem) it’s important to remember that everyone makes mistakes all the time. If someone made a mistake, so what? I believe it’s possible (and better for everyone) to be friendly and respectful when you correct someone’s mistake, rather than aggressive, sarcastic, dismissive, or insulting. Especially since, inevitably, some percentage of the time you go to correct someone, you’ll be mistaken. That’s only human.

Hope this helps clear things up.