Tesla cuts RAM on AI5 and AI6 chips to scale Optimus

By Splitfeed EditorialLast updated Oct 2, 2026, 05:00 UTC

On October 1 and 2, 2026, Elon Musk announced that Tesla has cut the memory specifications on two of its in-house chips to accelerate production of its Optimus humanoid robot. The AI5 chip had its RAM reduced by half, landing at 72 GB, while the AI6 chip was trimmed by a third to 144 GB. Musk framed the changes as a deliberate trade-off to remove a production bottleneck, and stated that the reductions would not affect Optimus performance in any meaningful way.

The significance of these cuts lies in what they reveal about Tesla's manufacturing situation. Cutting memory by as much as half is not a minor revision; it suggests that sourcing or producing higher-memory configurations was constraining how many chips Tesla could put into Optimus units. The company appears to be prioritizing volume, getting more robots out the door faster, over the raw specification numbers it had originally targeted. Whether the performance claim holds up in real-world deployment is the core technical question that remains open, since memory capacity in robotics chips tends to affect how much a system can process simultaneously, particularly for complex tasks that require holding large models or sensor data in working memory.

None of the three camps have posted public reactions to this story yet. The Pro-AI camp would typically treat Musk's assurance at face value, arguing that Tesla's engineering teams would not ship a degraded product and that scaling Optimus production faster is a net positive for robotics progress. The Anti-AI camp would likely be skeptical of the performance claim, noting that halving memory on a chip designed for an AI-driven robot is a meaningful reduction and that downplaying it serves a PR purpose ahead of production milestones. The Middle Ground camp would probably call for independent benchmarking before accepting either position, pointing out that the practical impact depends heavily on the specific workloads Optimus is running and how Tesla has optimized its software stack around the new constraints.

The clearest near-term signal to watch will be Tesla's Optimus production figures and any technical disclosures about real-world robot performance. If Tesla publishes deployment numbers or demonstration results that show no regression in capability, that would support Musk's claim. Any credible third-party analysis of the chip architecture under the new memory configuration would also help settle the dispute.