Elon Musk admits diverting Tesla’s AI chips to his other companies

Elon Musk admitted to diverting Nvidia’s latest AI chips originally intended for Tesla to his latest startup xAI and social media company X.

Confirming a CNBC report on Tuesday, Musk explained the move by saying that the chips, used to train neural networks, were better used elsewhere at the time.

“Tesla had nowhere to send the Nvidia chips to power them, so they would have just sat in a warehouse,” he posted.

This may not seem all that controversial for an electric vehicle maker, but it comes at a time when the CEO has gradually moved away from the goal of more than tenfold increasing annual car sales to the promised 20 million mark versus to Porcelain’s low-cost competition.

Instead, Musk has been methodically rebranding Tesla as an artificial intelligence and robotics company, pledging to invest $10 billion this year alone to beef up its AI training and inference computing. In April, he revealed plans to increase chip procurement to the equivalent of 85,000 Nvidia H100s by the end of this year just for AI training, that is, not including inference. It had around 35,000 at the end of the first quarter.

The timing of the revelation could also be damaging as Musk awaits the outcome of a tense June 13 vote on his record pay package, dubbed “the largest in human history,” whose previous approval (in 2018) was declared invalid. by a Delaware court because of governance failures. Tesla shareholders could fear they will bear the brunt if Musk sees the chips as more needed at xAI, SpaceX or X.

Musk raised fears that he could breach his fiduciary responsibility to Tesla shareholders after threatening earlier this year to further drive AI development out of Tesla if investors did not give him enough shares to increase his stake from 12% to 25%. % (another 9% held indirectly in the form of options granted as compensation is in dispute due to the Delaware ruling).

These concerns were exacerbated after his startup xAI began hiring employees, including from Tesla, allegedly because they were looking to leave the robotics and artificial intelligence startup anyway.

Musk’s conflicting interests

On Tuesday, Musk continued to say that the expansion of the electric vehicle maker’s site in Texas to house a new data center is almost complete, after the CNBC report cited warnings of possible additional delays in shipments of the intelligence chips. artificial reference from Nvidia to Tesla.

“This will house 50,000 H100s for FSD training,” the CEO said, a figure that explains the gap in the remaining chips to be purchased this year. He then added: “I cannot overstate the difficulty of getting 50,000 H100s trained as a coherent system. No company on Earth has been able to achieve this yet.”

It’s unclear why Tesla ordered the chips in the first place if they didn’t have a data center infrastructure in place that could use them at the time, and Musk didn’t elaborate.

One reason could be that Nvidia’s flagship H100 training chips (so advanced that the US banned their export to China) are in such demand that it couldn’t afford to risk giving up a potential delivery slot allocated by Jensen Huang’s company.

In the past, however, it has complained about not getting as many Nvidia GPU clusters as it needed, explaining that its planned Dojo supercomputer using custom silicon designed by Tesla was little more than a hedge against future delays.

It’s also unclear whether diverting some of Tesla’s chips actually hurt the automaker in any material way. The company says it invested $1 billion to more than double its computing capacity in the first quarter from the previous three months through the end of December.

In late March, Musk noted in a post that Tesla was no longer limited by AI training computing. This could suggest that the period where you had infrastructure capacity up and running but not enough chips to run it is over.

It’s possible that no harm was done to the company, and given the size of Musk’s financial interest, it would seem logical that he wouldn’t want to damage his only proven profitable effort.

However, his admission of diverting AI chips from Tesla to his other companies highlights the complicated nature of running a publicly traded company alongside several private companies, especially when they compete for the same resources. Musk sees no conflict of interest and does not hesitate to transfer resources as he sees fit.

During the test of Musk’s pay package, the CEO confirmed that he sent 50 of Tesla’s so-called volunteer engineers to help him on Twitter. James Murdoch, Tesla’s chairman on the board’s audit and governance committees, could not provide even a ballpark figure when asked about the decision, even though his testimony was provided only weeks later.

Leave a Comment