Your previous comment was clearly not constraining itself to the realm of AI chips / FSD. Regardless:
How do you measure Dojo, xAI, and Grok being failures?
Dojo is training models for FSD, which is in operation today. The chips in Tesla cars are also taped in-house / vertically integrated, a lot of which is presumably shared investment with the chips needed for the Dojo training side specifically.
Separating xAI from Grok (in a list like this) is kinda weird, but seems that Grok is actually a very capable LLM, even if it is not best-in-practice. Even if not beating out the top 3 labs, it certainly seems to be in the top 5 by most metrics. The xAI Colossus data centers are real AI training/inference infrastructure that were stood up in record time, and they are now selling capacity to other LLM providers. Is that a failure?
https://finance.yahoo.com/news/tesla-shuts-down-dojo-ai-2219...
Dojo is shutdown dude. Literally doing nothing but rotting. After spending $Billions on a custom chip, Tesla is now just buying NVidia chips and wasting their Dojo efforts entirely
Team dismantled. Operations closed. Game over.
---------
A similar event is playing out with xAI. Operations are transferring over to SpaceX. Out of money, out of time. They're going to try to grab more money (possibly illegally) across Musks enterprises but there's some severe legal questions on the legality of these moves.
It's already sketchy as all hell that xAI bought Tesla's GPU fleet. Now SpaceX is buying it up under doubious circumstances.
Whatever is going on with xAI / Twitter is now clear. It's not possible for it to stand on its own two feet and needs external investors to continue to survive.
Ok, but Dojo clearly has a spiritual successor in the form of: whatever Tesla is currently using to train FSD models? Like... Tesla kills teams all the time. Sounds like many were absorbed into the rest of the org. Seems like a bit of a nothing-burger without the follow-up that Tesla is shutting down FSD ambitions in general (or similar).
Yeah. It's called NVidia GPUs.
IE: Tesla spent over a $Billion building a team and custom building Dojo chips only to fire the entire team, throw away their chips and buy NVidia.
Are they replacing their onboard chips with NVidia chips?
AFAICT Tesla has increased investment in their own chip ambitions. Dojo specifically being cut is a massive failure if you are motivated to believe that, but otherwise it looks more like a subset of custom chip work that didn't pay off and the company has appropriately cut their losses. The onboard "AI2/3/4/5" chips are clearly the bigger investment (they can be found on every Tesla vehicle) and do not seem to be going anywhere at the moment.
We are just a few months away from the 10 year anniversary of Elon's coast-to-coast full self driving promise made in 2016.
If your idea of Elon Musks AI success is to lean on THAT side of Tesla... well, good luck with that.
Tesla is one of 3 companies currently operating a driverless taxi service.