Tesla reduces AI chip memory to scale Optimus robot production

Tesla is reducing the RAM capacity of its next-generation AI chips to facilitate the high-volume production of its Optimus humanoid robots and to lower hardware costs. Elon Musk announced that the AI5 chip’s memory will be cut from 144GB to 72GB of LP5, while the subsequent AI6 chip will see a one-third reduction from 216GB to 144GB of LP6.

Musk stated that the decision was necessary to address potential global DRAM supply bottlenecks that could arise if Tesla attempted to scale to billions of robots with higher memory specifications. He argued that the reduction would have a “negligible effect” on Optimus performance, asserting that memory bandwidth is a more significant limiting factor than total storage capacity.

The AI5 chip, which completed tape-out in April 2026, is intended primarily for Optimus and training supercomputers rather than vehicles, as the existing AI4 chip will continue to power Tesla’s cars. Tesla plans high-volume production of the AI5 chip for 2027. Manufacturing for these chips is being split between Samsung, using a 2nm process, and TSMC.