Elon Musk told Nvidia to send thousands of GPUs to X and xAI, which were originally reserved for Tesla.
In internal Nvidia emails seen by CNBC, staff also said that Musk's first-quarter earnings call comments on Tesla's GPU investment “conflicts with bookings."
Nvidia declined to comment, Musk did not respond to requests. On X (formerly Twitter), Musk said that "Tesla had no place to send the Nvidia chips to turn them on, so they would have just sat in a warehouse. The south extension of Giga Texas is almost complete. This will house 50k H100s for FSD training."
In the earnings call in April, Musk said that Tesla currently operates 35,000 H100 GPUs, and expected to increase that to 85,000 by the end of this year. On X a few days later, he said that Tesla would spend $10 billion this year “in combined training and inference AI.”
This, Nvidia memos said, did not align with what they had sold Tesla. Instead, the company swapped reservations for more than $500 million in GPUs to X.
“Elon prioritizing X H100 GPU cluster deployment at X versus Tesla by redirecting 12k of shipped H100 GPUs originally slated for Tesla to X instead,” an Nvidia memo from December said.
“In exchange, original X orders of 12k H100 slated for Jan and June to be redirected to Tesla.”
Another email, after the earnings call, said that the $10bn figure “conflicts with bookings and FY 2025 forecasts," and referenced drastic job cuts at the electric vehicle company, which could further delay the H100-based supercomputer at its Texas Gigafactory.
Musk has said that autonomous driving is key to Tesla's future - but to achieve that, it will need enormous GPU capabilities. These moves could slow such a transition, but may help with cash flow in the short term as the company's finances continue to struggle.
The celebrity CEO is also currently trying to convince shareholders to reinstate the largest executive rewards package in corporate history, a $56bn compensation deal that was voided by a Delaware judge.
Should it be reinstated, Musk would have a 20.5 percent stake in Tesla. If it isn't, he said that he “would prefer to build products outside of Tesla.”
He has also previously said that he was “uncomfortable growing Tesla to be a leader in AI & robotics without having ~25 percent voting control. Enough to be influential, but not so much that I can’t be overturned."
A number of AI researchers have left Tesla for xAI.
In an X post after the CNBC report was published, Musk said: "Of the roughly $10B in AI-related expenditures I said Tesla would make this year, about half is internal, primarily the Tesla-designed AI inference computer and sensors present in all of our cars, plus Dojo.
"For building the AI training superclusters, Nvidia hardware is about 2/3 of the cost.
"My current best guess for Nvidia purchases by Tesla are $3bn to $4bn this year."
This week, he also claimed that xAI would deploy a 300,000 B200 GPU supercomputer by next summer, and would soon launch an 100,000 H100 GPU data center. Neither of those projects have been independently verified.
xAI is believed to currently rent around 16,000 Nvidia H100 GPUs from Oracle Cloud. It also uses Amazon Web Services and spare capacity at X data centers, CNBC reports.
