Samsung to Pilot Tesla AI5 Chips on 2nm to Test Yield

According to Korean media reports, Samsung Electronics plans to activate its 2nm process line at the Taylor wafer fab in Texas, enabling simultaneous production of Tesla's next-generation AI5 and AI6 chips. During Tesla's Q3 earnings call, CEO Elon Musk confirmed that two companies will jointly manufacture the AI5 processor — revealing Tesla's new dual-foundry strategy involving Samsung and TSMC.

The AI5 chip, designed in-house by Tesla, serves as the central AI processor for its autonomous driving systems and humanoid robots. Currently, Samsung produces Tesla's AI4 chips using 4nm to 7nm process nodes, while the AI5 was initially expected to adopt TSMC's 3nm (N3AE) technology, with Samsung handling backend packaging and partial testing. However, as Tesla revises its foundry plans, Samsung is now expected to begin pilot production of AI5 on its 2nm GAA process by 2026 at the Taylor fab — a move that could help the company validate its 2nm yield and stability ahead of schedule.

Industry analysts note that TSMC still holds an advantage in 3nm process maturity and cost efficiency, while Samsung's entry into 2nm demonstrates strong technological ambition. Yet, with yield rates still stabilizing, cost competitiveness remains a challenge for Samsung's advanced nodes.

According to foreign media, Samsung's wafer foundry business is projected to post a loss of about KRW 5 trillion in 2024, but with improving 3nm yields and early ramp-up of the 2nm line, the division is expected to return to profitability by 2026.

Experts believe that if Samsung achieves stable mass production of Tesla's AI5 chips, it could strengthen its bid for AI6 manufacturing leadership and boost its position in negotiations with other high-performance computing clients. Moreover, Tesla's AI processors may eventually expand into humanoid robotics and data center applications, deepening the strategic collaboration between the two tech giants.