Musk transferred the thousands of AI chips reserved for Tesla to X and xAI

According to foreign news reports, Elon Musk claimed that he could develop Tesla into a leader in artificial intelligence and robotics, and he said that this goal will require a large number of expensive Nvidia processors to build its infrastructure.

In April, during Tesla's first-quarter earnings call, Musk said that the electric car company would increase the number of active units of Nvidia's flagship AI chip, the H100, from 35,000 to 85,000 by the end of the year.

A few days later, he also wrote in an article on X that Tesla would invest $10 billion in comprehensive training and inference AI this year.

But emails written by Nvidia's top brass and widely circulated within the company suggest that Musk has exaggerated Tesla's purchases to shareholders.

The letter from Nvidia employees also indicates that Musk has transferred a large number of AI processors reserved for Tesla to his social media company X (formerly Twitter).

Tesla shares fell 1% in pre-market trading on the news.

By ordering Nvidia to let the privately held X overtake Tesla, Musk has delayed the automaker's receipt of more than $500 million graphics processing units (GPUs) for months, which could lead to further delays in the installation of a supercomputer that Tesla claims needs to develop self-driving cars and humanoid robots.

An Nvidia memo in December said Elon was prioritizing the deployment of the X H100 GPU cluster at X over Tesla, moving the 12,000 H100 GPUs originally planned for Tesla to X. In exchange, orders for 12,000 H100 GPUs, originally scheduled for January and June deliveries, will be transferred to Tesla.

At the end of April, Nvidia sent an update email saying that Musk's comments on Tesla's first-quarter earnings report conflicted with order volume, while his comments on $10 billion in AI spending posted on X in April also conflicted with order volume and fiscal 2025 forecasts.

The email referred to news of Tesla's ongoing mass layoffs and warned that the layoffs could lead to further delays in Tesla's H100 program at Gigafactory Texas.

The new information in the email read by CNBC highlights the escalating conflict between Musk and some angry Tesla shareholders who question whether the billionaire CEO is meeting his obligations to Tesla, while also running a series of other companies that require his attention, resources and a lot of money.

A spokesperson for NVIDIA declined to comment on the matter. None of Musk, nor representatives of X and Tesla, responded to requests for comment.

Critics say Musk is only a part-time CEO of Tesla, which is the source of the vast majority of Musk's wealth. Musk is also the CEO of aerospace company SpaceX and the founder of brain-computer interface startup Neuralink and tunneling company The Boring Co. He also owns X, which he acquired for $44 billion at the end of 2022, when it was still called Twitter.

In 2023, he launched his own AI startup, xAI.

X and xAI are closely linked. In a November post about X, Musk wrote: X Corp investors will own a 25% stake in xAI.

In addition, CNBC understands that xAI uses part of the capacity of X's data center to run part of the training and inference of the large language model behind its chatbot, Grog.

Musk described Grok (originally named Truth GPT) as a politically incorrect chatbot with a rebellious personality and could be a competitor to OpenAI's ChatGPT and other generative AI services.

While Musk is handling multiple businesses, Tesla shareholders also have reason to be worried. The company is facing a troubling decline in sales, in part due to an aging EV lineup and increased competition.

The company's reputation in the U.S. has also suffered, according to the Axios Harris Poll 100 survey, which blamed part of the decline on Musk's antics and political rhetoric.

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

Musk didn't discuss EV sales or Tesla's ongoing massive restructuring, instead encouraging investors to focus on future products that he has promised for years but have yet to materialize. These products include artificial intelligence software that turns existing cars into self-driving cars, specialized robotaxis that can make money for car owners, and driverless transportation networks.

Musk said on an earnings call in April: "If someone doesn't believe Tesla will solve the problem of self-driving, I don't think they should be an investor in the company." We will, and we're doing it.

To achieve this, he said, Tesla needs a lot of Nvidia's GPUs that are dedicated to AI training and workloads. The supply of these chips is limited due to the surge in demand from companies such as Google, Amazon, Meta, Microsoft, OpenAI, and others.

Nvidia, which is currently the world's third-largest company by market capitalization with a market capitalization of $2.8 trillion, said it was struggling to keep up with demand. Nvidia CEO Jensen Huang said on an earnings call in May that customers, whether cloud service providers or companies developing AI models, are consuming all the GPUs on the market.

Previously, the chipmaker reported its third consecutive quarter of revenue growth of more than 200%.

Huang also said on the February earnings call that Nvidia will do its best to distribute fairly and avoid unnecessary allocations, adding that why allocate something when data centers aren't ready.

Referring to customers already using Nvidia's next-generation Blackwell platform, Huang mentioned xAI, as well as the world's six largest tech companies, as well as Tesla, in a May conference call.

Musk likes to brag about his infrastructure spending at both companies.

At Tesla, Musk pledged to build a $500 million Dojo supercomputer in Buffalo, N.Y., and build an ultra-dense, water-cooled supercomputer cluster at the company's factory in Austin, Texas. The technology will have the potential to help Tesla develop the computer vision and AI needed for robots and self-driving cars.

xAI is competing with companies like OpenAI, Anthropic, Google, and others in generative AI product development.

According to Nvidia's internal email in February, Musk is also seeking to build the world's largest GPU cluster in North Dakota, with some capacity to be operational in June.

The memo describes Musk's authorization to provide all 100,000 chips to xAI by the end of 2024. It states that the LLM behind xAI Grok relies on Amazon and Oracle cloud infrastructure, while X provides additional data center capacity.

The Information previously reported on some details of the xAI data center vision.

On May 26, xAI announced the closing of a $6 billion funding round, led by many investors who had previously funded Musk's acquisition of Twitter. The company was founded in March 2023, but Tesla did not disclose its establishment at the time, and Musk did not publicly introduce the startup until four months later.

Although Musk has said for years that Tesla is a leader in artificial intelligence, he said in his X post in January of this year that he wants to have more control over the company before pushing it further.

In the post, he said: "Without about 25% control of the vote, it would be difficult for me to develop Tesla into a leader in artificial intelligence and robotics. It's enough to make an impact, but not so much that I can't overturn my decision.

Tesla's latest proxy filings show that Musk owns 20.5% of the company's outstanding shares, a figure that includes options Musk received in an unprecedented CEO compensation package in 2018. A Delaware court has ordered the compensation to be revoked. Post-trial proceedings are ongoing and may be appealed.

In a January post, Musk said he would prefer to make products outside of Tesla if he can't meet his desired ownership goals. He's already doing that at xAI.

Musk's comments at the time angered some longtime bulls, including Leo Koguan, the company's largest retail shareholder, and Ross Gerber of Gerber Kawasaki, who described Musk's demands as blackmail.

Joel Fleming, a securities litigator at Equity Litigation Group, said Musk's conflict of interest was evident by allowing his private company to get ahead of Tesla in procuring critical hardware.

"When someone like Musk acts as a trustee for multiple companies at the same time, the law recognizes that this creates conflict." Fleming said. "If you have a fiduciary duty to two or more companies competing on the same thing, you may end up transferring business opportunities from one company to another."

Fleming, who often represents investors in shareholder disputes with public companies, said that in such cases, other executives are best qualified to make decisions, while those with conflicts of interest should abstain.

Fleming said: Historically, this is not the path that Musk himself has chosen.

Musk is not shy about mixing corporate resources between his companies.

For example, after acquiring Twitter, Musk recruited dozens of Autopilot software engineers and other technical and administrative staff from Tesla to help him overhaul the company. Some employees even work for two of Musk's companies at the same time.

At xAI, Musk has also poached employees from Tesla, including machine learning scientist Ethan Knight, and at least four other former Tesla employees who worked on Tesla's Autopilot and big data projects before joining the startup.

A former Tesla supply chain analyst revealed to CNBC that Musk has always seen his companies as an extension of himself personally and believes he can do whatever he wants with them. This includes Tesla's acquisition of SolarCity in 2016, when he was the company's chairman and largest shareholder. The analyst spoke on condition of anonymity due to the discussion of sensitive issues.

However, the person said that the move of a large number of chips from Tesla to X is a bit extreme, given the scarcity of Nvidia's technology. The decision means the automaker is willing to give up valuable time that could have been spent building supercomputer clusters in Texas or New York and advancing the models behind its self-driving software and robots.