Tesla AI6.5 Reportedly Before Foundry Switch: Weibo Rumor Brings Intel Into Play Instead of TSMC

Key Takeaways

• Tesla AI6.5 foundry shift to Intel remains unverified rumor
• Intel aims to attract major external customers with 18A/14A nodes
• US policy pressures favor domestic chip production for critical AI hardware
• Samsung secured $16.5 billion Tesla AI6 fab deal in Texas
• Multi‑sourcing would add design complexity for Tesla’s AI accelerators

Pulse Analysis

Tesla’s in‑house AI accelerator program has become a cornerstone of its autonomous‑driving and robotics ambitions. The first generation, AI4, powers the current Full Self‑Driving (FSD) hardware, while AI5, internally codenamed “Helios,” is already in volume production through TSMC and Samsung. The next step, AI6, is slated for a $16.5 billion fab partnership with Samsung’s new Texas plant, and an intermediate AI6.5 variant is rumored to be in development. These chips are expected to handle not only vehicle perception but also data‑center inference and Dojo‑related workloads, making their manufacturing source a strategic decision.

Intel’s foundry division has been courting external designers to validate its 18A and upcoming 14A processes, which feature RibbonFET, PowerVia and High‑NA EUV technologies. Securing a high‑profile customer such as Tesla would provide a powerful reference for the nascent node and help the company overcome lingering yield concerns that have hampered earlier external orders. However, adapting a large AI accelerator to Intel’s design‑rules, PDKs and packaging ecosystem is far more complex than a typical graphics‑card fab switch. Success would demonstrate Intel’s ability to compete with TSMC’s dominant advanced‑logic capacity.

The rumor surfaces amid a broader U.S. push to localize critical semiconductor production, with incentives encouraging companies to shift high‑performance workloads to domestic fabs. For Tesla, a multi‑sourcing strategy that includes Intel could mitigate supply‑chain risk and align with policy goals, but it also adds engineering overhead. TSMC’s 67 percent market share and Samsung’s newly funded Texas facility mean that any diversion of volume to Intel would reshape the competitive landscape. Even if the leak proves unfounded, it underscores how AI hardware decisions are now intertwined with geopolitics, security and the future of advanced‑node foundries.