Tesla Chip Orders to Boost Samsung's US Fab

The partnership between Tesla and Samsung dates back to 2019 for the 14-nanometer HW3 chip, with the current HW4 using a 5-nanometer process. This new long-term deal for the AI6 chip runs through 2033 and is valued at a minimum of $16.5 billion, initially for 16,000 wafers per month. Tesla is now reportedly requesting an additional 24,000 wafers per month for its AI6 chip, which would bring total production to 40,000 wafers. The chip will be built on Samsung’s 2-nanometer process, which is projected to deliver a 12% performance increase and a 25% reduction in power consumption compared to their previous node. A key architectural shift with the AI6 is its unified design for both training in data centers and inference in vehicles and Optimus robots. This eliminates the complex step of adapting models trained on different hardware, like Nvidia GPUs, to Tesla's own in-vehicle silicon, potentially accelerating development cycles. Production will be centered at Samsung's new facility in Taylor, Texas, though initial samples may come from South Korea. The Taylor plant, backed by a $17 billion initial investment and a $6.4 billion U.S. CHIPS Act grant, has faced delays, with full mass production now anticipated in 2026 or early 2027 instead of 2024. This deal is a significant victory for Samsung's foundry business, which holds an 8% global market share compared to TSMC's 67%. Securing a high-volume, leading-edge client like Tesla helps validate its 2nm process technology and builds a crucial anchor customer for the delayed Texas fab. The expansion is part of a broader push, fueled by the CHIPS Act, to increase the U.S. share of global semiconductor manufacturing from its current 12% back towards 30%. This is intensifying competition for advanced-node capacity and engineering talent in the U.S. as TSMC expands in Arizona and Intel invests nationwide.