Elon Musk has announced that Tesla is restarting work on its Dojo supercomputer project. Specifically, the renewed focus is on Dojo 3, the next generation of Tesla’s in-house AI training system, now that the design of its AI5 chip is in “good shape.”
Dojo, first unveiled in 2021, is Tesla’s custom-built supercomputer designed to process vast amounts of data from its car fleet – primarily through video and sensor inputs – to train AI models for Full Self-Driving (FSD) software, autopilot software, and applications for the company’s humanoid robot “Optimus.”
The project faced significant setbacks last year when Tesla paused Dojo development in mid-2025, with Musk calling it an “evolutionary dead end” and shifting resources to other AI chip designs.
At the time, Musk explained that dividing efforts between “distinct training” and “inference chips” didn’t make sense, leading to a team restructuring and increased dependence on companies like Nvidia and AMD for some of their computing needs.
In a January 18 X post, the Tesla CEO wrote, “Now that the AI5 chip design is in good shape, Tesla will restart work on Dojo3.”
Now that the AI5 chip design is in good shape, Tesla will restart work on Dojo3.
If you’re interested in working on what will be the highest volume chips in the world, send a note to [email protected] with 3 bullet points on the toughest technical problems you’ve solved.
— Elon Musk (@elonmusk) January 18, 2026
A single AI5 unit reportedly provides performance comparable to Nvidia’s “Hopper,” while two chips combined match the Blackwell class, all while consuming “much less power.”
“Solving AI5 was existential to Tesla, which is why I had to focus both the teams on that chip and I’ve personally spent every Saturday for several months working on it. This will be a very capable chip. Roughly Hopper class as single SoC and Blackwell as dual, but it costs peanuts and uses much less power,” Musk added.
Samples of the chip are expected later in 2026, with high-volume production scheduled for mid-2027.
Musk says that Dojo 3 will be the first Tesla supercomputer to use entirely in-house hardware, incorporating AI5, AI6, or even AI7 chips, without any external GPUs.
Looking ahead, Musk described an ambitious, fast-moving roadmap, with new chips planned roughly every nine months after AI5.
The next generation, AI6, is already in early development and is allegedly expected to deliver about twice the performance of AI5, powering both Optimus robots and data centers.
AI7, potentially integrated into Dojo 3, is envisioned for “space-based AI computing,” hinting at potential future collaborations with SpaceX’s Starlink network for “orbital data” processing.
Musk is now actively recruiting for the Dojo team, while predicting that Tesla’s chips could become the “highest volume AI chips in the world” due to their potential use in millions of cars and billions of future Optimus robots.
