Elon Musk ordered Nvidia to ship thousands of AI chips reserved for Tesla to X and xAI

Elon Musk aims to make Tesla a leader in AI and robotics, which will require expensive Nvidia processors to build its infrastructure.

On Tesla’s first-quarter earnings call in April, Musk stated that the electric vehicle company will increase the number of active H100s — Nvidia’s flagship artificial intelligence chip — from 35,000 to 85,000 by the end of the year. He also stated in a post on X a few days later that Tesla would spend $10 billion this year “in combined training and inference AI.”

However, emails written by Nvidia senior staff and widely circulated within the company indicate that Musk presented an exaggerated picture of Tesla’s procurement to shareholders. Correspondence from Nvidia employees also indicates that Musk diverted a large shipment of AI processors reserved for Tesla to his social media company X, formerly known as Twitter.

Tesla shares fell as much as 1% on the news Tuesday morning.

Musk pushed back the automaker’s receipt of more than $500 million in graphics processing units, or GPUs, by months by ordering Nvidia to allow privately held X to skip the line ahead of Tesla, likely adding to delays in setting up the supercomputers Tesla claims it needs to develop autonomous vehicles and humanoid robots.

“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,” a December Nvidia memo stated. “In exchange, original X orders of 12k H100 slated for Jan and June to be redirected to Tesla.”

According to a more recent Nvidia email from late April, Musk’s comment on the first-quarter Tesla call “conflicts with bookings” and his April post on X about $10 billion in AI spending also “conflicts with bookings and FY 2025 forecasts.” The email referred to Tesla’s ongoing, drastic layoffs and warned that headcount reductions could cause additional delays with a “H100 project” at Tesla’s Texas Gigafactory.

The new information from the emails, obtained by CNBC, reveals an escalating conflict between Musk and some agitated Tesla shareholders who question whether the billionaire CEO is fulfilling his obligations to Tesla while also running a number of other businesses that require his attention, resources, and significant capital.

A spokesperson for Nvidia declined to comment on this story. Musk and representatives from X and Tesla did not respond to requests for comment.

Critics claim Musk is only a part-time CEO of Tesla, the company that accounts for the vast majority of his wealth. Musk is also the CEO of SpaceX, the founder of Neuralink, and The Boring Co., a tunneling venture. He also owns X, which he bought for $44 billion in late 2022, when it was still called Twitter. He founded his AI startup, xAI, in 2023.

X and xAI are inextricably linked. In a November X post, Musk stated that “X Corp investors will own 25% of xAI.” CNBC has also learned that xAI uses some capacity in X data centers to run some of its training and inference for the large language models behind its chatbot, Grok.

Musk pitched Grok, formerly known as Truth GPT, as a politically incorrect chatbot with “a rebellious streak” and a potential competitor to OpenAI’s ChatGPT and other generative AI services.

Tesla shareholders have reason to be concerned as Musk manages his various ventures. The company is experiencing a troubling sales decline, which is being driven in part by its aging electric vehicle lineup and increased competition. Its reputation has also suffered in the United States, according to the Axios Harris Poll 100 survey, which attributed some of the decline to Musk’s “antics” and “political rants.”

Tesla’s stock price has fallen 29% this year.

Rather than discussing EV sales or Tesla’s massive restructuring, Musk has urged investors to focus on future products that he has promised for years but has yet to deliver. This includes AI software for transforming existing cars into self-driving vehicles, dedicated robotaxis that can generate revenue for their owners, and a driverless transportation network.

“If somebody doesn’t believe Tesla’s going to solve autonomy, I think they should not be an investor in the company,” Musk said during the April earnings call. “We will, and we are.”

He claims that Tesla will need a large number of Nvidia GPUs specialized for AI training and workloads in order to get there. High demand from Google, Amazon, Meta, Microsoft, OpenAI, and other companies has led to limited supply of these chips.

‘Consuming every GPU out there.’
Nvidia, the world’s third-most valuable company with a market capitalization of $2.8 trillion, has stated that it is struggling to meet demand. Customers “are consuming every GPU that’s out there,” Nvidia CEO Jensen Huang said on an earnings call in May, after the chipmaker reported more than 200% revenue growth for the third quarter in a row.

On an earnings call in February, Huang also stated that Nvidia strives to “allocate fairly and avoid allocating unnecessarily,” adding, “Why allocate something when the data center isn’t ready?”.

On the May call, Huang mentioned xAI, along with six of the world’s largest tech companies and Tesla, as customers who are already using Nvidia’s next-generation Blackwell platform.

Musk frequently boasts about his companies’ infrastructure spending.

Tesla CEO Elon Musk has promised to build a $500 million “Dojo” supercomputer in Buffalo, New York, as well as a “super dense, water-cooled supercomputer cluster” at the company’s Austin factory. The technology could help Tesla develop the computer vision and LLMs required for robots and self-driving cars.

Musk is also looking to build “the world’s largest GPU cluster” in North Dakota, with some capacity available in June, according to an internal Nvidia email from February.

The memo described a “Musk mandate” that would make all 100,000 chips available to xAI by the end of 2024. The LLM for xAI’s Grok relies on Amazon and Oracle cloud infrastructure, with X providing additional data center capacity.

The Information previously reported on xAI’s data center ambitions.

On May 26, xAI announced that it had closed a $6 billion financing round led by many of the same investors who had funded Musk’s Twitter acquisition. The company was founded in March 2023, but Tesla did not announce its formation until four months later, when Musk publicly introduced the startup.

Conflicts of Interest
While Musk has claimed for years that Tesla is a leader in artificial intelligence, he stated in a January post on X that he would like more control over the company before going any further.

“I am concerned about Tesla becoming a leader in AI and robotics without having ~25% voting control.” “Enough to be influential, but not so much that I can’t be overturned,” he wrote in his post.

According to Tesla’s most recent proxy filing, Musk owns 20.5% of the company’s outstanding shares, including options granted to him as part of his unprecedented 2018 CEO pay package. A Delaware court has ordered that compensation be rescinded. Post-trial proceedings are ongoing and open to appeal.

If he is unable to reach his desired ownership level, Musk stated in a January post that he “would prefer to build products outside of Tesla.” He is already doing that at xAI.

Musk’s comments at the time irritated some long-time bulls, including the company’s largest retail shareholder, Leo Koguan, and Gerber Kawasaki’s Ross Gerber, who called his demand “blackmail.”

According to Joel Fleming, a securities litigator at Equity Litigation Group, Musk’s conflicts of interest are highlighted by allowing his private companies to outbid Tesla in the procurement of critical hardware.

“When you have someone like Mr. Musk who is a fiduciary to multiple companies, the law recognizes this creates conflict,” Fleming stated. “If you owe fiduciary duties to two or more companies that are competing over the same things, you may end up channeling corporate opportunity away from one company to another.”

Fleming, who frequently represents public company investors in shareholder disputes, believes that in such cases, other executives are best positioned to make decisions, while those who are conflicted should abstain.

“That has not historically been the path that Mr. Musk has chosen for himself,” Fleming stated.

Musk has not been shy about sharing corporate resources between his companies.

For example, after buying out Twitter, Musk enlisted dozens of Tesla Autopilot software engineers and other technical and administrative employees to assist him in making significant changes at the company. Some employees even work for two Musk companies simultaneously.

Musk has also attracted former Tesla employees to xAI, including machine-learning scientist Ethan Knight and at least four others who worked on Autopilot and big data projects before joining the startup.

A former Tesla supply chain analyst, who asked not to be identified in order to discuss sensitive matters, told CNBC that Musk has always viewed his companies as extensions of himself and believes he can do whatever he wants with them. This includes Tesla’s 2016 acquisition of SolarCity, where he was chairman and a major shareholder.

However, the person stated that redirecting a large shipment of chips from Tesla to X is extreme, given the scarcity of Nvidia’s technology. The decision means that the automaker willingly gave up valuable time that could have been spent expanding its supercomputer cluster in Texas or New York and developing the models that underpin its self-driving software and robotics.