Dojo Is Officially Dead. Elon’s $500B Supercomputer Now Lives in the Same Cloud as Harambe and Hope

Well, pour one out for Dojo.

The AI supercomputer Elon Musk once pitched as a self-evolving brain… built to devour mountains of real-world video, train itself, and give Tesla a greek godlike edge in the robotaxi race? Yeah, it’s been disbanded. Shut down. Thrown in the trash like a crushed Bud Light can at a tailgate. And if you’re wondering whether that $500 billion in enterprise value Morgan Stanley wildly predicted (without a shred of evidence) still exists... it’s probably hanging out with Harambe and Vine up in the clouds.

If you remember, back in 2021, Elon Musk introduced Dojo during Tesla’s first AI Day (feels like yesterday, doesn’t it?). The system came with its own custom-built D1 chip, plenty of hype, and big promises… to make Autopilot more capable, deliver true Full Self-Driving, and eventually power Tesla’s robot army to handle everyday tasks, like vacuuming your house.

The mission had every analyst in America nodding their heads in unison: Elon and his buddies would take raw video data from Tesla’s fleet, feed it to Dojo, and in return, get a self-driving AI that improves continuously… eventually powering an autonomous ride-hailing network and laying the foundation for Tesla’s Optimus humanoid robot ambitions. In other words, this wasn’t another random Elon pet side project (remember his flamethrower?) it was actually central to Tesla’s pitch as an AI company, not just a carmaker. “We are NOT a car company, do you understand me?!?”.

Then came Morgan Stanley’s now infamous note, projecting that Dojo could add up to $500 billion to Tesla’s valuation by accelerating the rollout of robotaxis and AI-powered services. That estimate, to put it gently, aged like milk. For context, $500 billion would be more than the entire market cap of Visa or ExxonMobil today. But hey, in a ZIRP-fueled world of AI optimism, nobody wanted to be the one throwing cold water (especially Elon’s way).

Well fast forward to this morning, and it’s all over. Not paused. Not restructured. Shut down. The team is gone. Dojo lead Pete Bannon (no relation to Steve) is out. About 20 former engineers left to launch DensityAI, a startup now aiming to do exactly what Tesla couldn’t (because nothing says "this is fine" like your team immediately starting a competitor). The rest were reassigned to Tesla’s data center group.

(Source: CNBC)

Why? Because apparently, running two separate AI chip efforts was “not a good use of resources.” Musk now wants to focus on Tesla’s newer AI5 and AI6 chips (designed with Samsung) and just lean more heavily on external partners like Nvidia and AMD. Which is funny, because Dojo was supposed to replace Nvidia. But now? Tesla's deploying 50,000 Nvidia H100s in a massive new cluster called Cortex… because, let’s be honest, when you need to crank out AI training, you go back to the dealer you secretly hate.

And another thing, I’m sure there’s already dozens of thinkpieces written by Tesla cult members who are writing this off as a longshot from the beginning… but I’m not gonna sugarcoat it. This was a major L. Dojo was the symbol of Tesla’s claim to being an AI and robotics company first, and an automaker second. It was supposed to justify Musk’s claims that Tesla was massively undervalued relative to its long-term potential. Instead, we got demo robotaxis with human “safety supervisors” and a humanoid robot that walks like it owes money to both knees.

And while Musk insists that Tesla and xAI (his other AI company) are pursuing separate goals, it’s hard to ignore the awkward overlap. Tesla’s losing senior engineers faster than a fintech startup post-IPO, and some of the company’s brightest minds are now building chips elsewhere. Look, even Elon admitted in early 2024 that Dojo was a “long shot.” But the market? It treated Dojo like a sure thing. Like Tesla had a patent on Skynet. Like robotaxis were about to start paying off your student loans.

Instead, it’s back to the basics: build the chips, buy the GPUs, and try to deliver FSD before 2040 (or maybe by the time we’re all forgetting to take our meds).

Next time someone says “this AI project could add $500 billion in value,” maybe double-check they didn’t mean “negative $500 billion in delusions.” Elon, you magnificent meme king… thanks for the ride. And as for DOJO, say hi to Haramabe for me.

At the time of publishing this article, Stocks.News holds positions in Tesla and Exxon as mentioned in the article.