Key Takeaways
- Tesla’s AI5 chip production will be split between Samsung in Texas and TSMC in Arizona
- Elon Musk confirmed Tesla will maintain its partnership with Nvidia for AI training
- The company aims for an oversupply of AI5 chips for use in vehicles, robots, and data centers
- Tesla’s in-house chip design could deliver 10x better performance per dollar
Tesla’s Dual-Source Chip Strategy
Tesla CEO Elon Musk has confirmed that the company’s upcoming AI5 chip will be manufactured by both Samsung in Texas and TSMC in Arizona. During Tesla’s third-quarter earnings call, Musk revealed the company plans to produce more chips than it immediately needs, stating: “Our goal is to have an oversupply of AI5 chips.”
Any excess chips not used in Tesla vehicles or robots will be deployed in the company’s data centers, creating a flexible supply chain for Tesla’s growing AI needs.
Nvidia Partnership Continues
Despite developing its own processors, Tesla will continue relying on Nvidia’s GPUs for AI model training. Musk clarified: “We’re not replacing Nvidia, but we use both in combination.” The company currently possesses computing power equivalent to 81,000 Nvidia H100 chips.
Tesla previously used Nvidia’s Drive chips before switching to its own processors in 2019. The AI5 chip, first announced in 2024, represents the latest version of Tesla’s Autopilot hardware designed for self-driving functions.
Strategic Manufacturing Decisions
The decision to split production between Samsung and TSMC strengthens Tesla’s U.S.-based chip supply chain. In July, Samsung disclosed a $16.5 billion chip contract with an unnamed client—later confirmed to be Tesla. Musk also confirmed that the next-generation AI6 chip will be produced by Samsung.
The AI5 chip features a unique “half-reticle” size, making it smaller than most existing AI designs from competitors like Nvidia and AMD.
Performance Advantages
Musk explained that Tesla’s in-house chip design simplifies optimization since “Tesla only has to satisfy one customer—ourselves. That makes the design job radically easier.”
Engineers removed older components including GPUs and signal processors from the AI5 design, potentially delivering “the best performance per dollar for AI—maybe by a factor of 10.”
Industry Context
This move positions Tesla alongside tech giants like Google, Amazon, and Microsoft, all developing custom AI chips to reduce Nvidia dependence. Currently, only Apple and Tesla design their own silicon for both devices and data centers.
Meanwhile, Musk’s AI startup xAI remains a major Nvidia customer and is building a new supercomputing center in Memphis, Tennessee, powered by Nvidia’s latest Grace Blackwell chips.
