Micron’s 200GB Robot-Memory Forecast Meets Tesla’s Lower-Memory Optimus Plans

Micron’s forecast of more than 200GB of DRAM and several terabytes of storage per humanoid robot is best read as a possible future configuration, not a universal specification. The comparison with Tesla’s Optimus chips makes the uncertainty clear: memory needs depend on the robot’s design and how much AI processing it handles onboard. Micron sees physical AI as a potential significant source of demand by the end of the decade, but the forecast alone does not show when robot orders will materially affect the memory market. 171920

Micron’s figure is a projection, not a robot standard

Micron’s benchmark comes from Level 4 and higher autonomous vehicles, where memory content typically exceeds 200GB and storage reaches multiple terabytes. The company has pointed to humanoid robots as a possible extension of this physical-AI trend. But applying the vehicle figure to robots is a forward-looking estimate, not evidence that every robot will ship with that configuration. 171920

The distinction between DRAM and storage matters, too. DRAM is working memory; storage is a separate capacity measure. A forecast for both does not mean a robot needs hundreds of gigabytes of DRAM for every onboard task, or that all robot designs will use the same balance of memory and storage.

Tesla’s chip plans show how much configurations can differ

Tesla initially said it would reduce planned memory on its AI5 chip from 144GB to 72GB and on AI6 from 216GB to 144GB, citing supply, production volume and cost. Elon Musk later said AI5 would be raised to 96GB; the reported AI6 plan remained 144GB. These figures describe memory planned for the chips—not a confirmed total DRAM-and-storage specification for a complete Optimus robot. 1615

Musk said Tesla expected the cuts to have a negligible effect on Optimus performance, arguing that memory bandwidth was a bigger limitation than total capacity. That is Tesla’s stated view, not an independent demonstration of how the changed specifications will affect the finished robot. 16

The comparison therefore has limits: Micron’s figure is a broad projection for a future robot, while Tesla’s figures concern particular AI chips. Still, the announced plans are a useful reminder that a robot’s memory configuration is a design choice, not a fixed number that can be inferred from Micron’s vehicle benchmark.

What Micron’s market signals do—and don’t—tell us

Micron reported record fiscal 2026 results, and the company said more than 75% of its expected fiscal 2027 output was already committed. It also expected memory and storage supply-demand conditions to be tighter in fiscal 2027 and 2028 than in fiscal 2026. Those signals point to a constrained market, but they do not establish that humanoid robots are already a major source of demand. 334917

Micron’s LPDDR6 sampling for physical-AI markets suggests it is preparing products for potential future applications. Sampling is not the same as broad deployment, and it does not establish how many robots will use the parts or how much memory each will contain. 18

When could robots become a significant memory buyer?

Micron has described physical AI as a potential significant memory and storage demand driver by the end of the decade. That makes humanoid robots a longer-term possibility, rather than a proven explanation for the tight supply outlook in 2027 and 2028. The available information does not pin down a point when robot demand will become significant: that depends on how quickly deployments grow and how much memory manufacturers put in each machine. 1719

For now, the most grounded takeaway is that robot memory demand could be substantial, but its scale is unsettled. Micron’s 200GB-plus estimate sketches one possible high-memory future; Tesla’s revised chip plans show why actual configurations may be lower or otherwise different.