Elon Musk’s xAI Teams With Nvidia, Dell for AI Supercomputer

Elon Musk's AI startup xAI is partnering with Dell, Nvidia and Supermicro to construct what it claims will be the world's largest supercomputer. 

The project, announced in early June, will provide the startup with a “gigafactory of compute,” its own dedicated infrastructure to power its training efforts and reduce its current reliance on cloud services from Oracle and several of X’s data centers.

News of Dell’s involvement broke after CEO Michael Dell posted a photo of server racks on X (formerly Twitter), saying his company was building one of its “AI factories” as part of the supercomputer’s infrastructure.

Dell's AI factory concept offers specialized AI infrastructure services, utilizing liquid-cooled servers and Nvidia's new Blackwell GPUs to support high-intensity AI workloads.

The Dell CEO said the AI factory would be used by xAI to power Grok, the startup’s rebellious chatbot designed to compete with OpenAI’s ChatGPT.

Musk confirmed that Dell will be assembling half of the racks going into the supercomputer.

The company tasked with building the other half is Supermicro. Musk confirmed their involvement in a response to another X user, simply writing “SMC” — Super Micro Computer.

Supermicro did not confirm it’s working with xAI. 

Musk revealed earlier this month that xAI plans to acquire approximately 300,000 new Blackwell B200s from Nvidia, though this won't happen until "next summer."

Related:Musk xAI Plans World's Largest Supercomputer In Memphis

Musk said given the pace of AI hardware advancements, he did not see it worth investing in Nvidia’s current flagship H100 GPUs. 

Nvidia, which recently became the most valuable company in the world, has adopted a new yearly release schedule, launching new GPUs every year including updated versions of its Blackwell and new Rubin chips.

XAI recently raised $6 billion and despite being a one-year-old, is valued at $24 billion.

Musk’s AI startup has also been receiving shipments of Nvidia chips previously earmarked for Tesla. 

Some 12,000 H100 GPUs destined for Tesla were diverted to X, according to CNBC. Musk had stated during the company’s first-quarter earnings call that Tesla would be acquiring around 85,000 Nvidia H100s.

Confirming the move, Musk said Tesla “had no place to send the Nvidia chips to turn them on, so they would have just sat in a warehouse.”