Musk says Tesla ‘not about to replace Nvidia’ as EV maker builds chips
Tesla CEO Elon Musk listens to U.S. President Donald Trump speak to reporters in the Oval Office of the White House in Washington, DC, on May 30, 2025.
Kevin Dietsch | Getty Images
Tesla’s The upcoming artificial intelligence chip, which he calls AI5, will be manufactured by Samsung in Texas and TSMC in Arizona, CEO Elon Musk told investors on Wednesday.
On the electric vehicle maker’s third-quarter earnings call, Musk said Tesla was pursuing “overproduction” and that any chips not used in cars or robots could be used in his company’s data centers.
“Our clear goal is to have an oversupply of AI5 chips,” Musk said.
Tesla has been used before Nvidia’s It uses chips in its vehicles but abandons them in favor of its own processors in 2019. Musk said Tesla will continue to use graphics processing units from Nvidia, which dominates the artificial intelligence market, for training its models.
“Obviously we are not going to replace Nvidia, but we are using both together,” Musk said. Tesla announced Wednesday that it has the computing capacity equivalent to 81,000 of Nvidia’s H100 chips.
Musk’s comments offer a new perspective on Tesla’s artificial intelligence chip strategy.Apple Engineer Peter Bannon left the company earlier this year. Bannon previously led chip design for Tesla and led the development of Dojo, a supercomputer aimed at helping the company develop its driverless technology.
The AI5 chip was first announced at Tesla’s shareholders meeting in 2024. This is the latest version of the car company’s Autopilot hardware, which processes the signals necessary for self-driving features. In July, Samsung announced that it was secured. $16.5 billion chip contract with an unnamed customer – Musk later confirmed that was Tesla.
At that time, Musk said that the AI5 chip would be produced by TSMC, and its successor, the AI6 chip, would be produced by Samsung. The AI5 chip will be produced at both foundries’ U.S. facilities, it said on Wednesday. It said it would fit on “half reticle,” meaning the chip is half the size of Nvidia’s “full reticle” AI chip designs, and Advanced Micro Devices.
Nvidia CEO Jensen Huang holds a motherboard while speaking at the Viva Technology conference dedicated to innovation and startups at the Porte de Versailles exhibition center in Paris, France, on June 11, 2025.
Gonzalo Fuentes | Reuters
Hyperscalers like: Google, Amazon And Microsoft They are all investing in their own AI chips to offer an alternative to Nvidia, and some AI experts say specialized chips may be more cost-effective or faster for certain tasks.
Other than Tesla, only Apple designs its own silicon and uses it both in its products and in data centers for AI services.
Musk said that because Tesla is the sole customer for AI5, chip designers were able to focus on the company’s needs and eliminate legacy parts of the chip that were slowing it down.
“Tesla only needs to meet one customer’s needs,” Musk said. “This greatly simplifies the design work and means we can remove a lot of complexity from the chip.”
He said Tesla’s design team removed the chip’s old GPUs, signal processors and other parts and predicted the chip could have “maybe 10 times more” the best performance per dollar for AI.
“Nvidia has done a great job of tackling almost incredibly difficult requirements,” he said. “But in our case we expect radical simplicity.”
Musk’s artificial intelligence startup xAI, which does business with Tesla, has become a major Nvidia customer. The company is currently building a supercomputing facility in Memphis, Tennessee, the second largest in the region, based on high-end Nvidia Grace Blackwell chips.
— CNBC’s Lora Kolodny contributed to this report.
WRISTWATCH: Tesla shares tumble after Musk omits mention of robotaxi fleet
