During his financial report to investors in April, Elon Musk said that Tesla would increase the number of active H100 artificial intelligence chips from NVIDIA from 35,000 to 85,000 by the end of the year. However, the billionaire’s statements contradict his actions, as shown by emails between senior NVIDIA employees, which state that Musk redirected a significant batch of chips that were being prepared for Tesla to the X company.

According to CNBC, NVIDIA emails from December last year, which they obtained, say that Elon Musk redirected 12 thousand chips worth more than $500 million that were supposed to be sent to Tesla to the X. Instead, X’s order for the same number of chips will be sent to Tesla, but later.

This decision delays for months Tesla’s ability to customize the supercomputers that the company needs to develop a fully autonomous driving system and humanoid robots, in particular because such processors are now in high demand. NVIDIA’s letters also state that Elon Musk gave Tesla investors a very overstated picture of the company’s chip purchases.

Further, the NVIDIA employee emails say that the billionaire’s April post on X that Tesla will spend about $10 billion on training and artificial intelligence this year conflicts with bookings and forecasts for fiscal year 2025.

Other information that CNBC was able to obtain from the correspondence highlights the conflict between Musk and some of the company’s investors, who question whether the billionaire can really fulfill all the responsibilities of Tesla’s CEO while he has many other companies, such as SpaceX, Neuralink, X, xAI, and The Boring Co.

In response to the CNBC story, Elon Musk wrote in X that Tesla has nowhere to send these chips, which means they would just be in storage. In addition, he said that the company’s new gigafactory in Texas is almost ready and that it will house 50,000 H100 processors. He also stated that $3 to $4 billion of the previously promised $10 billion in AI spending would be spent on NVIDIA chips.