Tesla AI Chip Delayed by Samsung Foundry Slip
Tesla's AI6 chip production is delayed by roughly six months because Samsung's 2nm foundry schedule slipped. Mass production is now expected in Q4 2027, creating ripple effects for other projects and customers. This delay highlights the need for robust float in schedules for long-lead items and frequent check-ins with specialty suppliers to mitigate potential cascading delays.
The delay in Tesla's AI6 chip is not isolated, impacting other Samsung 2nm foundry clients like AI chip startup DeepX, who planned to use the node for their DX-M2 processor. This processor was designed for on-device generative AI, supporting models up to 100 billion parameters at 5 watts, with mass production initially slated for Q2 2027. DeepX now anticipates quality testing in Q3 2027 and sales in Q4 2027. This setback stems from a postponed multi-project wafer (MPW) run, a cost-effective prototyping method where multiple companies share a single wafer. If one customer's schedule shifts, it affects all sharing the wafer. Tesla's AI6 chip delay highlights how major players can influence production timelines for smaller startups. Samsung's 2nm process was crucial for their foundry division's profitability goals, aiming for 2 trillion won in profit this year, with the AI6 chip as a key revenue driver. The delay raises concerns about Samsung's ability to compete with TSMC in advanced process nodes. Elon Musk had previously stated that both Samsung and TSMC would manufacture AI5 and AI6 chips, using slightly different versions of the same design. The AI6 chip is central to Tesla's plans for autonomous vehicles, Optimus robots, and AI infrastructure, with a contract running through 2033. Tesla initially secured 16,000 wafer starts per month with discussions to increase to 40,000. The AI6 delay also compounds existing issues, as volume production of the AI5 chip has been pushed to mid-2027, potentially forcing the Cybercab to launch with older hardware.
