Elon Musk has confirmed that both Tesla and SpaceX will continue to purchase Nvidia hardware in large volumes, even as Tesla advances its proprietary fifth-generation AI chip, known as AI5. This dual strategy highlights a complementary approach where in-house chips are designed for specific edge-computing tasks while Nvidia's infrastructure remains critical for large-scale AI model training.
Tesla's AI5 chip, set for early production via the Terafab project launching on March 21, is optimized for real-time, edge computing applications. It is designed to power Tesla's Full Self-Driving (FSD) software, the humanoid robot Optimus, and the Robotaxi platform, processing data directly within vehicles and robots rather than relying on remote data centers. The company is already working on a sixth-generation chip, AI6, with an ambitious goal to move a new chip design into full production annually.
Despite this internal push, Musk expressed admiration for Nvidia and its CEO Jensen Huang, emphasizing that Nvidia's hardware is indispensable for training complex AI models. Tesla's proprietary chips, including the current AI4 and upcoming AI5, are built to work alongside, not replace, Nvidia's ecosystem. This reliance underscores the continued dominance of Nvidia's computing infrastructure in the AI sector.
On the software front, a wider rollout of Tesla's Full Self-Driving Supervised update is expected within the next few weeks, expanding its assisted driving features to more users. The structural integration of artificial intelligence across Musk's ventures is also deepening, following SpaceX's acquisition of xAI last month. Musk now refers to the combined entity as "SpaceX AI," signaling a unified effort to scale computing resources across Tesla, SpaceX, and related projects ahead of a potential SpaceX IPO later this year.
