When you purchase through links on our articles, Future and its syndication partners may earn a commission.

Tesla and xAI, Elon Musk’s companies, will bring training computing capacity worth $10 billion online by the end of this year, as observed Sawyer Merrittco-founder of TwinBirch and Tesla investor. And yet, that likely means both companies will be somewhat behind the timeline set by Elon Musk.

Elon Musk and his companies have recently made active announcements about AI supercomputers. So we are indeed talking about huge investments.

In July, xAI began AI training using the Memphis Superclusterwhich is expected to integrate 100,000 liquid-cooled systems H100 GPU. This system requires gargantuan power, consuming at least 150 MW, with the 100,000 H100 GPUs alone representing around 70 MW. The total cost of the system is unknown, although the GPUs alone would cost around $2 billion (if purchased at $20,000 each), and typically AI GPUs are half the cost of the entire system.

At the end of August, Tesla unveiled its Cortex AI Clusterequipped with an impressive 50,000 Nvidia H100 GPUs and 20,000 of Tesla’s own Dojo AI wafer-sized chips. The Dojo cluster is expected to form Tesla’s fully autonomous driving (FSD) capability, so this machine is strategically vital to the company.

As for costs, we are talking about two billion for the machine based on the H100 and at least one billion for the Dojo supercomputer. This billion could be underestimated as Dojo machines are entirely custom designed. For example, each Dojo D1 cabinet consumes over 200 kW (to put things in context, each B200 NVL72 (the cabinet is expected to consume 120 kW) and therefore requires a completely custom Cooling Distribution Unit (CDU) and power supply, which significantly increases its cost.

Finally, in early September, xAI began exploiting its Colossus supercomputer, which already integrates 100,000 H100 GPUs and should add 50,000 H100 and 50,000 H200 GPU in the coming months. This giant AI supercomputer also costs billions.

XAI and Tesla have likely announced spending well over $10 billion on AI hardware this year. Of course, it will be a while before all of these AI servers are installed and brought online, so we can only guess at the total cost of working AI hardware installed by both companies in 2024.

But the ironic thing about this huge spending is that it appears to be somewhat behind the ambitious plan Elon Musk laid out last April, when he said Tesla alone would spend $10 billion on hardware. ‘AI this year.

“Tesla will spend about $10 billion this year on a combination of training and inference AI, the latter of which is primarily aimed at automotive,” Musk said. wrote in a post. “Any company that doesn’t spend at that level, and doesn’t do it efficiently, can’t compete.”

While Tesla’s Cortex AI cluster is likely an expensive undertaking that will likely become more expensive over time if the company decides to install more Dojos or more Nvidia-based machines, we doubt it will cost much more than, say , 5 billion dollars. As for the costs of AI inference hardware in cars, we can’t imagine that the AI ​​computing hardware in vehicles that will be produced by Tesla this year will cost $5 billion.