Tesla AI5 Chip Enters Trial Production at Samsung's Texas Fab

Tesla’s next-generation silicon is moving from design to fabrication. Samsung has begun trial production of Tesla’s AI5 chip at its new Taylor, Texas foundry, shifting the effort into physical validation as the company verifies manufacturing yields ahead of high-volume output.

According to a report from Seoul Economic Daily, Samsung recently began wafer fabrication on its advanced 2-nanometer node at the Taylor facility (via Drive Tesla Canada). Although full operations were initially planned for later in the year, Samsung accelerated prototype production to handle a surge of custom silicon orders. A key driver is Tesla, which secured a $16.5 billion contract to have its next-gen chips manufactured at the site.

A 5X Upgrade in Computing Power

The move to AI5 is more than an incremental refresh for Tesla’s self-driving hardware. It parallels the leap from Hardware 3 to Hardware 4, where HW3 provided only 15% of the memory bandwidth available on HW4/AI4. Looking ahead, AI5 is expected to outclass AI4 by delivering roughly 5x the raw performance.

Current vehicles use two AI4 chips operating in parallel on a single board. Elon Musk has said a single AI5 chip can deliver five times the compute power of two AI4 chips combined. That performance density is expected to help reduce power consumption while giving neural networks room to scale.

Where AI5 Heads First

Prototype runs began after Musk confirmed earlier this spring that the AI5 design was complete and had exited the tape-out stage, enabling Samsung to translate the digital design into test patterns on silicon.

Samsung is targeting the end of 2026 to finish yield verification, with mass production at the Texas fab slated to begin in 2027. The early start to trial production supports that schedule. However, buyers should not expect AI5 in a new Model 3 or Model Y soon. Musk has noted that AI5 will debut in Optimus robots and in AI training clusters at Tesla data centers long before reaching consumer vehicles.

Tesla maintains that the existing AI4 platform has sufficient headroom to achieve unsupervised autonomy. Driverless, steering wheel-less Cybercabs are already operating commercially without AI5 — though they use a more powerful FSD computer than consumer builds — allowing Tesla to prioritize the newest silicon for its humanoid robot program and AI training. Eventually, Tesla plans to move to a nine-month development cycle for future chip iterations, with designs roughly outlined up to AI9.

What About AI4+?

For older vehicles, the path differs. While Tesla says AI4 can handle both the latest full FSD builds and upcoming unsupervised capabilities, it has acknowledged that HW3 cars will require hardware upgrades to achieve full autonomy. The company has announced an upgraded AI4+/AI4.5 computer with more memory that is widely viewed as the retrofit solution for HW3, though owners are still awaiting official details on production and availability.

With trial production underway in Texas, Tesla’s longer-term hardware roadmap is now materializing in silicon.