Even if that were the case, that can't go on forever. Even if it's literally just a matter of who has enough data, at some point a competitor, maybe with state-level support, will have "enough" data to come after you with their own self driving option. Perhaps now Tesla is at 90% (just say, it's just an example) and Mercedes (again say) is at 10%. One day Tesla may be 100% and Mercedes is at 30%, and the German government, for one, will be shoveling cash into the furnace the whole time if the German auto industry is in existential danger. The gap can only close.

Meanwhile once you reach full self driving, you're stuck at diminishing returns. Doubling your data won't double that: at best you could make a very marginal improvement to safety stats or range economy. Maybe with a lot of lobbying you could swing a slightly higher Tesla-only speed limit, but double self driving isn't a thing. Everything else in the car is commodity stuff like sound systems and seat heaters then anyone with a car factory and a team of designers and a copy of the Bosch Automotive Handbook can do if they want to. Probably you could get some network effects from the "Rent your Tesla out" app, but again, that's an option for anyone with a team of designers once the self-driving part is done and they'll be in the disruption phase and aggressively undercutting as best they can.

This also assumes that in 2029 you can't just use 80% (again... just a random number) synthetic training data and leapfrog nearly all of those hard-won physical miles. The costs of doing that will plummet too, so the training cost moat will only shrink as well.

Meanwhile Tesla will be ossifying as gigantic trillion dollar behemoths tend to do and accruing costs like pensions, and management entrenching will be in its third decade and so on.

Is Tesla worth a lot? Undoubtedly. Is it that much? Questionable. But then the market disagrees, so what do I know (but also the market was off by 30% by it's own estimation so how much does it really know?)

I'm not sure I agree. In our (very very) hypothetical scenario, it seems that self driving would play out similar to how software incumbents in general are very hard to displace.

You could use the same argument for search, operating systems, or public cloud as well? "Why couldn't a more nimble or well funded company come in and commoditize {software industry}"

I'm not sure, but we only have two major app stores, we only have three big public clouds, we only have three companies making major operating systems for consumer devices, two search engines, I could go on.

I think that marginal benefit that comes from first mover and scale really makes it hard to catch up in the world of software. Big tech is not afraid to use rapidly rising stock to retain talent, which makes it very hard for startups to acquire comparable talent unless VCs are willing to throw gobs of money at companies, that then go to exorbitant comp. (and you see this in the bull market run of 2015-2022).

Even if that didn't ignore the scaling behavior of auto manufacturing, which it very badly does, that would only be a reason it might be winner-take-all, because it has a giant load-bearing “if”.

right, above I mentioned that it's too big of an if for me to invest. Hence why I think Tesla is a decent car that people should consider, but I'm not a shareholder.