Suspicion Over AI Chip X Supply for Tesla... Musk Refutes, "Will Spend 4-5 Trillion on Chip Purchases"

CNBC Reports "Tesla Sends Nvidia Chips to X"

Elon Musk, CEO of Tesla, clarified plans to invest 4 to 5 trillion won in purchasing NVIDIA AI chips, stating that Tesla's AI technology development this year will proceed as scheduled, in response to reports that NVIDIA AI chips ordered for Tesla were diverted to another of his companies, X.

On the 4th (local time), Musk posted on his social media platform X (formerly Twitter), "About half of Tesla's approximately $10 billion (about 13.738 trillion won) AI-related expenditure this year is internal costs, mainly used for Tesla-designed AI inference computers, sensors installed in vehicles, and Dojo (a supercomputer for autonomous driving training)."

Elon Musk Tesla CEO [Photo by Yonhap News]

He added, "NVIDIA hardware for building AI training superclusters accounts for about two-thirds of that cost," and "The current estimate for Tesla's NVIDIA chip purchases is $3 billion to $4 billion (about 4.1214 to 5.4952 trillion won) this year."

This statement came following a report by the U.S. economic media CNBC that Musk directed AI chips ordered for Tesla to be supplied first to X.

In Tesla's Q1 earnings announcement in April, Musk revealed that Tesla would increase its purchase of NVIDIA's flagship AI chip, the H100 GPU, from the initially planned 35,000 units to 85,000 units. He also announced an investment of $10 billion in AI this year through X.

However, according to the CNBC report citing internal documents and emails from NVIDIA employees, Musk instructed that a significant portion of AI chips reserved for Tesla be shipped first to X.

According to a document prepared by a senior NVIDIA executive in December last year, Musk reassigned 12,000 H100 units originally intended for Tesla to X, prioritizing the deployment of X's H100 GPU cluster over Tesla's. Instead, X's orders for 12,000 H100 units scheduled for January and June this year were allocated to Tesla.

Additionally, emails from NVIDIA employees at the end of April this year indicated that Musk's remarks during Tesla's Q1 conference call and X's post about investing $10 billion in AI conflicted with NVIDIA's AI chip reservations. There was also a warning that the 'H100 project' at Tesla's Texas Gigafactory could face further delays.

Musk explained that this was not about prioritizing other companies over Tesla. Through X, he stated, "Tesla had no space to operate the NVIDIA chips, so they would have just been sitting in a warehouse," adding, "The southern expansion of the Texas Gigafactory is nearly complete and will accommodate 50,000 H100 units for FSD (Full Self-Driving) training."

CNBC pointed out that Musk managing multiple companies simultaneously, including Tesla, X, and SpaceX, could lead to conflicts of interest. It also noted that this aligns with Musk's past statements about developing AI externally if he cannot secure sufficient shares in Tesla. In January, Musk said on X, "Growing Tesla as a leader in AI and robotics without 25% voting rights (shares) is uncomfortable," and if he cannot secure 25% shares, he would "prefer to build products outside Tesla."

According to CNBC, Musk holds 20.5% of Tesla's shares, which includes stock options granted under a 2018 compensation package. However, minority shareholder Richard Tornetta filed a lawsuit seeking to invalidate the compensation package and has temporarily won, which may require Musk to return the stock options. Tesla's board has scheduled a reapproval of the compensation plan at the shareholders' meeting on the 13th.

This content was produced with the assistance of AI translation services.