Musk Says Tesla AI6 Chip Will Break Efficiency Records

Musk Says Tesla AI6 Chip Will Break Efficiency Records

Tesla is developing the silicon that it expects to underpin its autonomy plans. This week, Elon Musk shared takeaways on X from an engineering review of the AI6 chip, outlining expectations for the next-next-generation platform.

Musk posted on X: "Tesla AI chip design engineering reviews are so great! Team is awesome. Our AI6 chip might set a record for most amount of usable intelligence from a wafer when factoring in yield."

AI6 is being designed to span a broad ecosystem, including Full Self-Driving for commercial Robotaxis and consumer vehicles, Optimus humanoid robots, and space-based data center applications. While AI6 remains in the engineering and design phase, the upcoming AI5 chip has already been taped out and is planned for mass production in late 2027. With Musk targeting a nine-month development cycle for Tesla's future AI chips, AI6 is expected to enter production in the latter half of 2028.

Major Architectural Upgrades

The performance gains are set to be significant. The upcoming AI5 chip has been described as delivering 5 times the compute power of two AI4 chips, which power current Tesla vehicles. Musk has said AI6 will double the performance of AI5, indicating not only a generational jump but also an architectural shift in processing and memory management.

Starting with AI5, future hardware platforms will include much more memory, as the latest FSD models have been capping out the available memory on AI4.

To prevent processing bottlenecks, both AI6 and the mid-cycle AI6.5 refresh Musk mentioned earlier this spring will dedicate roughly half of their TRIP AI computation accelerators to SRAM. This ultra-fast onboard memory allows the processor to run complex AI calculations in a high-speed workspace without waiting on main system memory. For main memory, the AI6 design moves to faster LPDDR6, a step up from the LPDDR5 or LPDDR5X configurations expected on the next-gen AI5 chip architecture.

Production and Integration Plans

Manufacturing plans are already underway. Tesla is working closely with Samsung, which will manufacture AI6 in Texas at its new semiconductor facility. Beyond that $16.5 billion chip manufacturing deal, Tesla has also partnered with Intel and SpaceX on the TERAFAB AI chip project to vertically integrate semiconductor production.

Owners should not expect this new silicon in cars immediately. Musk has specified that Tesla's next-gen AI chips are not coming to vehicles first, with initial deployment planned for Optimus robots and the supercomputer clusters that train Tesla’s neural networks before scaling down to consumer cars. Musk already considers current AI4 setups powerful enough to beat human driving safety, giving the hardware team time to refine AI6 before it reaches public roads.