Elon Musk Says Tesla’s Revived Dojo3 Will Power ‘Space-Based AI Compute’

Elon Musk says Tesla is bringing back Dojo again but with a very different goal this time. Over the long weekend, the Tesla CEO revealed that the company plans to restart work on Dojo3, its once-abandoned third-generation AI chip. Unlike earlier versions designed to train self-driving systems on Earth, Dojo3 will be aimed at something far more ambitious: space-based AI compute.

The announcement comes just five months after Tesla effectively shut down its Dojo program. The company disbanded the Dojo supercomputer team following the departure of project lead Peter Bannon, and roughly 20 engineers left soon after to join DensityAI, a new AI infrastructure startup founded by former Dojo head Ganesh Venkataramanan along with ex-Tesla employees Bill Chang and Ben Floering.

At the time, Bloomberg reported that Tesla planned to lean more heavily on external partners using Nvidia and AMD for compute and Samsung for chip manufacturing rather than continue developing its own custom silicon. Musk’s latest comments suggest that strategy has shifted yet again.

In a post on X, Musk said the decision to revive Dojo was driven by progress on Tesla’s internal chip roadmap, noting that the company’s AI5 design is now “in good shape.”

AI5, manufactured by TSMC, was built to power Tesla’s Full Self-Driving software and its Optimus humanoid robots. Last summer, Tesla also signed a $16.5 billion deal with Samsung to produce its upcoming AI6 chips, which are expected to run Tesla vehicles and Optimus while also supporting large-scale AI training in data centers.

Now Musk is looking even further ahead.

“AI7/Dojo3 will be for space-based AI compute,” he wrote on Sunday, framing the resurrected project as a full-blown moonshot literally.

To pull it off, Tesla will need to rebuild the team it dismantled earlier this year. Musk used the same post to recruit engineers directly, writing:

“If you’re interested in working on what will be the highest volume chips in the world, send a note to AI_Chips@Tesla.com with 3 bullet points on the toughest technical problems you’ve solved.”

The timing is hard to ignore. At CES 2026, Nvidia unveiled Alpamayo, an open-source AI model for autonomous driving that directly challenges Tesla’s Full Self-Driving software. Musk responded by acknowledging how difficult the problem really is, saying that solving rare edge cases in driving is “super hard,” before adding, “I honestly hope they succeed.”

Musk isn’t alone in thinking the future of AI infrastructure may be off-planet. Several AI leaders have argued that Earth’s power grids are already stretched thin, making space-based data centers an attractive long-term solution. Axios recently reported that OpenAI CEO Sam Altman is also intrigued by the idea. Musk, however, has a unique advantage: he controls the rockets.

According to Axios, Musk plans to use SpaceX’s upcoming IPO to help finance a vision that involves launching a constellation of AI compute satellites using Starship. These systems would operate in near-constant sunlight, harvesting solar power around the clock.

Of course, massive challenges remain especially cooling high-power compute hardware in the vacuum of space. Still, Musk’s comments fit a familiar pattern: introduce an idea that sounds wildly impractical, then attempt to brute-force it into existence.

Whether Dojo3 becomes the backbone of orbital AI infrastructure or another abandoned experiment remains to be seen. But if nothing else, Tesla’s ambitions just got a whole lot more cosmic.