Tesla Revives Abandoned Dojo3 Chip for Space-Based AI Compute
Elon Musk announces that Tesla is restarting its previously scrapped Dojo3 project, shifting its focus from terrestrial self-driving training to off-planet AI processing. The company is actively recruiting engineers to rebuild the team it disbanded just five months ago.
Elon Musk reveals that Tesla is restarting work on its third-generation AI chip, Dojo3, just five months after the company disbanded the original team. Instead of training self-driving models on Earth, the revived project focuses entirely on space-based AI compute. Musk bases this strategic pivot on the strong progress of Tesla's in-house AI5 chip design, which is currently manufactured by TSMC to power the automaker's vehicles and Optimus humanoid robots.
Tesla is now actively working to rebuild the dismantled Dojo team by recruiting engineers directly on X to tackle what Musk describes as the highest volume chips in the world. This sudden reversal follows Tesla's previous decision to rely on Nvidia, AMD, and Samsung for its computing needs, including a massive $16.5 billion deal with Samsung to build AI6 chips. The timing of the announcement also coincides with Nvidia unveiling an open-source autonomous driving model at CES 2026 that directly challenges Tesla's FSD software.
By positioning Dojo3 as an off-planet computing solution, Musk aligns Tesla with a growing belief among tech executives that the future of data centers lies in space due to Earth's strained power grids. The ambitious project essentially acts as a moonshot for the electric vehicle maker. Musk uses the shift to address terrestrial limitations while continuing to push the boundaries of his interconnected companies' technological capabilities.