Key facts
- Elon Musk stated Tesla's upcoming AI6 chip will be the world's best edge computing chip.
- Tesla has taped out its next-generation AI5 processor, with samples expected later this year.
- The AI5 chip is designed for Optimus robots and supercomputer clusters, not vehicles.
- AI6 is targeted for tapeout in December 2026 and aims to double AI5 performance.
- AI6 will be manufactured by Samsung on a 2-nanometer process node in Texas.
- AI5 will be dual-sourced from TSMC in Arizona and Samsung in Texas.
Elon Musk has declared that Tesla's forthcoming AI6 semiconductor will be the world's leading edge computing chip, signaling the company's strategic pivot towards artificial intelligence and robotics. The announcement follows Tesla's tapeout of its AI5 processor, with initial silicon samples anticipated by year-end and high-volume production slated for mid-2027.
The AI5 chip, designed for real-time AI inference, will power Tesla's Optimus humanoid robot and its supercomputer clusters, but will not be used in vehicles, which will continue to utilize the current AI4 hardware. Musk indicated that AI5 offers a substantial performance leap over AI4, with a single chip comparable to an Nvidia H100 GPU for specific workloads, and a dual-chip configuration rivaling Nvidia's Blackwell-class processors, all while consuming less power and costing significantly less.
Musk also provided an aggressive roadmap for future chip development, aiming for new designs to reach volume production approximately every 12 months. The AI6 chip, targeted for tapeout in December 2026, is expected to double the performance of AI5 within the same physical footprint. This chip will utilize LPDDR6 memory and will be manufactured by Samsung at its new 2-nanometer fab in Texas. Tesla's AI5 chip will be dual-sourced, with production split between TSMC's Arizona facility and Samsung's Texas plant.