Tesla's AI5 Chip Reaches Tape-Out at Samsung Foundry on 2nm Process

Inside the AI5 Module

Tom's Hardware described the AI5 module shown by Musk in April as combining a relatively compact accelerator die — roughly half a reticle in size, based on Musk's earlier remarks — with 12 memory packages supplied by SK hynix. The publication noted that these appear to be standard GDDR6 or GDDR7 devices, though the exact specification was not confirmed in the coverage.

Tesla has positioned its custom AI silicon as central to its driver-assistance and autonomy ambitions, so a second qualified manufacturing source could help the company scale supply and reduce dependence on a single foundry. The reliance on leading-edge 2nm-class nodes also places Tesla among the earliest customers seeking to use both major foundries' most advanced processes for a high-volume automotive and AI product.

Why It Matters for the Global Chip Supply Chain

The development underscores continued competition between TSMC and Samsung at the cutting edge of semiconductor manufacturing. Samsung's use of its Taylor, Texas fab for the AI5 also highlights the growing role of U.S.-based advanced capacity, a theme that has featured in broader coverage of foundry investment and geopolitics.

Several details remain unconfirmed, including a firm production start date, output volumes, and precise chip specifications. The information disclosed so far originates from a single engineer's social-media post rather than a formal company announcement, and neither Tesla nor Samsung has issued detailed public statements described in the source material. As a result, the timeline and technical particulars may be clarified as the companies move toward volume manufacturing.