Tesla Cuts RAM on AI5 and AI6 Chips to Boost Production
Elon Musk says reducing memory by a third enables volume production for the Optimus robot and lowers unit costs.
Tesla is reducing the memory capacity of its upcoming AI5 and AI6 chips to facilitate higher production volumes, particularly for the Optimus humanoid robot. The move, announced by CEO Elon Musk, involves cutting RAM specifications to overcome supply constraints and lower manufacturing costs.
What Happened
Elon Musk shared the specifications change on X earlier this week, stating that Tesla cut RAM for the AI5 chip to 72GB of LPDDR5 and reduced AI6 memory by a third to 144GB of LPDDR6. Musk described the reduction as "the only way to get enough volume for Optimus production," noting that the change also significantly reduces per-unit costs. Following an initial plan to cut AI5 RAM in half, Musk adjusted the specification to 96GB, resulting in a one-third reduction from the original plan to avoid using the minimum RAM version of LP5. Musk stated that Tesla expects a "negligible effect on Optimus performance," arguing that memory bandwidth, rather than total capacity, is the limiting factor for these systems and that overall bandwidth remains unchanged.
Why It Matters
The decision reflects the intense pressure on the global memory supply chain due to surging AI demand. Micron has estimated that each humanoid robot could require approximately 200GB of RAM, meaning Tesla’s reduced specifications leave the company operating with less than half the memory industry analysts suggest might be necessary. Despite the cuts, the new chips represent a significant upgrade over current hardware; AI4 computers currently ship with 32GB of GDDR6, while Hardware 3 used 8GB of LPDDR4 per chip. Samsung has begun trial production of AI5 at its Taylor, Texas facility, with mass production targeted for 2027. Musk indicated that a single AI5 chip delivers five times the compute power of a full two-chip AI4 computer. However, the AI5 hardware will prioritize Optimus robots and Tesla’s AI training clusters before potentially appearing in passenger vehicles, with Musk maintaining that current AI4 hardware is sufficient for unsupervised autonomy.
The Bottom Line
Tesla is prioritizing production volume and cost efficiency for its next-generation AI chips by reducing RAM specifications, betting that memory bandwidth rather than capacity will sustain performance for the Optimus robot. While AI5 mass production is aimed for 2027, the hardware will not initially be deployed in consumer vehicles.
