Elon Musk said on Wednesday evening that SpaceX and Tesla will maintain large-scale purchases of NVIDIA Corporation chips while Tesla simultaneously pushes ahead with its own next-generation AI chip program.
In a post on the social network X, Musk described himself as a "huge admirer" of Nvidia and its chief executive, Jensen Huang. He confirmed that both SpaceX and Tesla will keep ordering Nvidia chips at scale.
Musk provided an update on Tesla's AI5, the automaker's in-house AI processor that remains under development. He said AI5 will perform "far above its weight" relative to expectations and that the architecture is primarily tuned for Tesla's Optimus humanoid robot and the robotaxi project. Musk had previously said the AI5 design was nearing completion, though outside reports indicate volume production has slipped to mid-2027.
Related to Tesla's broader chip ambitions, Musk referenced Terafab, a project intended to build AI chips. Musk said last week that Terafab is scheduled to launch on March 21. Separately, Tesla signed a $16.5 billion agreement with Samsung in mid-2025 tied to construction of its own AI chips, reflecting the company's push to integrate custom silicon.
At the same time Tesla continues to rely on Nvidia technology for AI workloads. The company previously dismantled its Dojo computing complex last year and has used Nvidia chips to meet AI requirements. Musk has repeatedly framed custom AI and robotics as critical growth avenues for Tesla amid two consecutive years of declining EV sales, positioning proprietary processors as strategic for Optimus and robotaxi ambitions.
Musk also touched on chip use across his other ventures. xAI - the AI startup recently merged into SpaceX - trained its flagship Grok model using Nvidia chips. The Colossus computing facility in Memphis that supports xAI primarily runs on Nvidia's H and Blackwell families of processors.
SpaceX itself said it plans to build AI data centers in space, and the company is preparing for a substantial public offering this year. Despite those plans and Tesla's own chip program, Musk reiterated that Nvidia hardware will continue to be procured at scale by both companies.
These comments underline an ongoing dual approach: continued reliance on established GPU suppliers for existing compute needs, paired with a strategic investment in bespoke silicon aimed at specific robotics and autonomy applications.