Tesla AI5 Chip to Use Samsung’s 2nm Process for Production

Tesla is preparing to begin production of its AI5 chip using Samsung’s 2nm process technology after completing the chip’s tape-out earlier this year. The new processor will power Tesla’s next-generation Full Self-Driving system and deliver a major increase in AI performance compared with the current HW4 chip.

Samsung will reportedly manufacture a large share of the Tesla AI5 chip at its Taylor facility in Texas. TSMC will also handle part of the production, although details about the process node used at its factories remain unclear.

Yonhap News and The Guru reported that Samsung plans to build the chip using its 2nm manufacturing process. A Samsung Foundry engineer also stated that the AI5 design had entered the tape-out stage and would later appear in Tesla’s latest products.

The chip includes a large central compute die surrounded by 12 DRAM modules, which appear to come from SK hynix. This design gives the processor more memory capacity and higher bandwidth for advanced AI workloads.

Elon Musk previously said the AI5 chip would deliver up to eight times more raw computing power and nine times more memory than HW4. Reports also suggest that it will offer around 2,500 TOPS of AI performance and up to 144GB of memory per chip.

Tesla plans to offer single-chip and dual-chip versions of AI5, with the more powerful configuration targeting performance similar to Nvidia’s Blackwell platform. High-volume production is expected to begin in late 2026 or early 2027, while Tesla has already started work on AI6 and Dojo 3.