• At a private party on Thursday, Elon Musk spoke boldly about Tesla's upcoming custom artificial intelligence hardware.
  • The remarks from Musk come almost two years after Tesla hired chip architect Jim Keller.
  • Tesla's current Autopilot autonomous driving hardware uses Nvidia graphics cards.

Tesla CEO Elon Musk talked up the company's work to develop custom hardware for artificial intelligence on Thursday during a talk at a company party for academic and industry researchers.

For the event in Long Beach, Calif., Musk brought with him Tesla's vice president of hardware, Jim Keller. Musk talked about the company's AI hardware in the course of providing information about what Keller, a former AMD chip architect, is working on.

"Jim is developing specialized AI hardware that we think will be the best in the world," Musk said, according to one person who was at the event, which was not streamed live.

The specialized hardware could one day be used inside Tesla vehicles to do the computing work necessary for autonomous driving. Currently, Tesla's Autopilot hardware system relies on graphics cards from Nvidia.

CNBC reported on Tesla's AI chip effort in September.

Several other automakers, including Ford and GM, are developing self-driving technology. Alphabet, which has developed custom chips that can be used instead of Nvidia graphics cards, operates the Waymo self-driving car division. Apple has pursued autonomous driving, and it has also developed custom silicon to handle AI workloads on mobile devices.

In the Thursday talk, Keller suggested that with custom hardware it's possible to boost efficiency, while Musk talked about power and cost advantages, according to one report about the event on Twitter.

tweet about the event

Musk also provided a projection for when fully automated driving will be available: two years.

tweet about the event

It's not clear, however, when Tesla would start selling cars with the custom hardware. Tesla did not immediately respond to a request for comment.