Deepening collaboration between Samsung and Tesla anticipated as Dojo project concludes
Tesla, the electric vehicle (EV) giant led by CEO Elon Musk, has announced a strategic pivot in its AI chip development, focusing on the AI5 and AI6 chips. This shift comes as the company ends its Dojo supercomputing project, signaling a deeper reliance on external chip partners.
The AI5 chip is a low-power inference accelerator intended for in-vehicle deployment to support Full Self-Driving (FSD) and robotics applications, with production expected to begin by the end of 2026. On the other hand, the AI6 chip is a high-performance training chip aimed at data center use and onboard inference for autonomous functions, emphasizing expansion capacity, high memory bandwidth, and mainstream AI framework compatibility.
In a significant move, Tesla has signed a substantial $16.5 billion deal with Samsung Electronics to produce the AI6 chips. While the AI5 chips are fabricated by TSMC, the AI6 chips are entrusted to Samsung for wafer manufacturing, reflecting Tesla’s strategic choice to outsource fabrication to specialized foundries while retaining in-house chip design control.
Samsung's role in manufacturing Tesla's AI6 chips is expected to grow, following the deal which spans through 2033 and involves the production of Tesla's next-generation AI6 chips at Samsung's plant in Taylor, Texas. Beyond manufacturing, Samsung is expected to expand its role, working closely with Tesla's design teams to jointly develop products.
This pivot represents a strategic streamlining to create versatile chips with unified software support, enhancing Tesla’s ability to integrate advanced AI across its EVs and robotics platforms efficiently. Elon Musk has stated that Tesla will not divide resources to pursue different AI chip designs, hinting at a potential convergence between Tesla's technology and that of its partners.
The news follows Musk's video call with Samsung's Chair Lee Jae-yong to discuss potential partnerships. The call implied that Tesla might be moving towards a partnership model for future AI chip development, with Tesla planning to increase its reliance on external technology partners, including Nvidia, AMD, and Samsung Electronics.
Lee Jae-yong has been in the US since July 29, with speculations growing that he may meet with Tesla's top executives during his stay. During the same call, Musk mentioned a potential convergence between Tesla's Dojo 3 and AI6 inference chip. However, the Dojo project aimed to build a custom supercomputer for training machine-learning models behind Tesla's Autopilot, Full Self-Driving systems, and Optimus humanoid robot, and its discontinuation marks a significant shift in Tesla's AI strategy.
Elon Musk described the disclosed amount as "just the bare minimum," implying the actual value could be higher. This strategic pivot during Tesla's most recent quarterly earnings call suggests that Tesla is aiming for a more collaborative approach in its AI chip development, signaling a promising future for both Tesla and its partners.
Business lies at the heart of Tesla's decision to partner with Samsung Electronics for the production of the AI6 chips, a move that underscores the company's growing reliance on technology in both its finance and business operations. The collaboration between Tesla and Samsung is expected to expand beyond manufacturing, with Samsung also working closely with Tesla's design teams to jointly develop products.
