Skip to Content
MIT Technology Review

Tesla is making its own AI chips

Elon Musk confirmed during Tesla’s most recent earnings call that the company is making its own computer chips for automated driving (Tesla currently relies on hardware from Nvidia). “We’ve been in semi-stealth mode on this basically for the last 2-3 years,” Musk said. “I think it’s probably time to let the cat out of the bag.”

Not so secret: In fact, Musk first revealed the project last December, by describing details during an AI conference. The effort was widely rumored after Tesla hired Pete Bannon, the engineer who oversaw development of Apple’s A5 chip, in 2016.

Better driving: It isn’t clear whether faster hardware will improve automated driving significantly. Many of the problems experienced so far have involved sensors and software safeguards rather than the machine learning code.

AI skillz: Making its own hardware for machine learning could, however, provide Tesla with a competitive advantage over companies that rely on off-the-shelf technology. It will certainly help Tesla differentiate itself among carmakers, and allow Musk to talk up the company’s AI expertise and capabilities.

Turbo boost: Musk said the new hardware would offer an order of magnitude improvement over Nvidia’s hardware (although he didn’t specify which generation). He also said that older Teslas could be retrofitted with the hardware.