Elon Musk says Tesla’s restarted Dojo3 will be for ‘space-based AI compute’

Elon Musk has announced plans to revive Tesla's Dojo3, the company's previously shelved third-generation AI chip project. Unlike its original purpose of training self-driving models on Earth, this new iteration will focus on what Musk describes as 'space-based AI compute.' This shift comes just five months after Tesla halted its Dojo initiative and disbanded the team behind the supercomputer following significant personnel changes. The departure of key figures, including Dojo lead Peter Bannon, and around 20 team members who joined the startup DensityAI, left Tesla's AI efforts in a state of uncertainty. Initially, Tesla had pivoted towards relying more on partnerships with Nvidia and AMD for computing needs, as well as Samsung for chip manufacturing, rather than pursuing the development of proprietary silicon. Musk's latest remarks indicate a strategic reversal. He emphasized the importance of the company’s in-house chip roadmap, stating that the design of Tesla's AI5 chip is 'in good shape.' This chip, produced by TSMC, is intended to enhance the performance of Tesla’s automated driving features and the Optimus humanoid robots. Furthermore, Tesla's recent $16.5 billion agreement with Samsung aims to produce AI6 chips that will power vehicles and support advanced AI training within data centers. The billionaire entrepreneur framed Dojo3 as a pioneering project for space-based AI. To make this vision a reality, Musk is actively seeking to rebuild the team, inviting engineers to express their interest via a direct call for talent. He wrote on X, '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 of this announcement is particularly significant. At CES 2026, Nvidia showcased an open-source AI model, Alpamayo, designed for autonomous driving, which poses a challenge to Tesla’s Full Self-Driving (FSD) software. Musk remarked on social media that addressing the rare edge cases in driving is 'super hard,' expressing hope for their success in the endeavor. Musk and other AI leaders are increasingly considering the potential of off-planet data centers, as Earth’s energy infrastructure faces increasing pressures. Reports have suggested that OpenAI's CEO Sam Altman also shares interest in the concept of orbital data centers. Musk's existing control over launch vehicles gives him a distinct advantage; he plans to leverage SpaceX’s upcoming IPO to fund his ambitious project of deploying a network of computing satellites, designed to operate in constant sunlight and harness solar energy around the clock. Despite the ambitious nature of establishing AI data centers in space, challenges remain, particularly in managing the cooling of high-performance computing within a vacuum. Musk's vision of 'space-based AI compute' aligns with his pattern of proposing seemingly audacious ideas and then striving to bring them to fruition.