Elon Musk confirmed that many Nvidia Corp. chips initially intended for Tesla’s electric vehicle production were diverted to X Corp. due to logistical challenges.
“Tesla had no place to send the Nvidia chips to turn them on, so they would have just sat in a warehouse,” the EV maker’s chief executive said in a post on X on Tuesday.
Musk’s post was in response to a CNBC report earlier Tuesday that cited a December memo from Nvidia last year saying that 12,000 of Nvidia’s H100 graphics processing units, its flagship artificial intelligence chip, meant for Tesla were redirected to X. Similar orders for X that were slated for deliveries in January and June this year were then shifted to the EV company, according to the CNBC report.
Nvidia didn’t immediately respond to a request for comment.
Tuesday’s post may bolster criticism that Musk’s push for strategy changes, such as focusing on robotics and driverless cars, would make Tesla’s traditional car business less of a priority. His AI plans have come under scrutiny after Musk threatened in January to take his ideas on advanced technology elsewhere if he isn’t given more ownership of the EV maker.
Musk holds 13% of Tesla shares directly, and about 21% if unexercised options are included. He has publicly requested 25% ownership to have more voting power in corporate matters. Tesla shareholders will meet June 13 for the company’s annual general meeting, where they will also vote on Musk’s pay package, recently valued at billions of dollars.
In the Tuesday post, Musk also said the extension works on Tesla’s Gigafactory in Texas are almost completed and will house 50,000 H100 chips. Musk added in a separate post that about half of the $10 billion in AI-related expenditures that Tesla is expected to make this year would be internal, primarily for the Tesla-designed AI inference computer and sensors present in all of its cars, as well as Tesla’s supercomputer dubbed Dojo.
“For building the AI training superclusters, Nvidia hardware is about two-thirds of the cost,” Musk said Tuesday, adding that Tesla would spend between $3 billion and $4 billion on hardware purchases from Nvidia this year.
Tesla has been working on its own supercomputer as part of efforts to develop driverless-car technology. During April’s first-quarter earnings call, Musk said Tesla would increase the number of active H100s from 35,000 to about 85,000 by the end of this year.
Write to Kimberley Kao at kimberley.kao@wsj.com