Musk says Tesla cut AI5 and AI6 chip memory for Optimus

By Splitfeed EditorialLast updated Oct 2, 2026, 03:10 UTC

On or around October 1 and 2, 2026, Elon Musk announced that Tesla had cut the memory specifications for two chips destined for its Optimus humanoid robot. The AI5 chip had its memory reduced by half, landing at 72 GB, while the AI6 chip saw a one-third reduction, bringing it to 144 GB. Musk framed both cuts as deliberate engineering decisions tied to scaling Optimus production rather than as concessions driven by supply shortages or cost pressures alone, and he stated directly that the performance of the Optimus robot would not suffer as a result.

The significance here lies in what the numbers reveal about the tension between manufacturing ambition and raw hardware capability. Cutting memory in half on a flagship AI chip is not a minor tweak; it is a substantial reduction that would ordinarily raise questions about what tasks the robot can handle simultaneously, how much it can hold in working memory during complex operations, and whether future software updates will be constrained by the lower ceiling. Musk's assurance that performance is unaffected is the central claim in dispute, because it is unverified by independent benchmarks or third-party testing at this stage. Whether Tesla achieved genuine algorithmic efficiency gains that justify the cuts, or whether the performance guarantee will hold as Optimus moves into more demanding real-world environments, remains an open question.

With no published reactions from the Pro-AI, Anti-AI, or Middle Ground camps yet, it is possible to sketch what each would typically argue. Pro-AI voices would likely treat Musk's framing at face value, pointing to the memory reductions as evidence that Tesla's software and model compression have matured enough to do more with less, which would be a genuine engineering achievement. Anti-AI commentators would typically be skeptical of the performance-unaffected claim, arguing that memory cuts of this scale almost always involve tradeoffs that only become visible under real operating conditions. Middle Ground observers would probably call for independent verification, noting that scaling production and maintaining capability are not inherently in conflict but that the burden of proof rests with Tesla to demonstrate both can coexist.

The clearest near-term signal to watch would be any public demonstration or third-party evaluation of Optimus units built with the revised AI5 specifications, as well as Tesla's projected production volume figures for Optimus in the coming quarters, which would indicate whether the scaling rationale is translating into actual output.