AI firm xAI cancelled a large server deal with Oracle

Generative AI company xAI has cancelled a deal to advance a server business deal with Oracle based on differing time and supply terms. After cancellation, xAI is pursuing its plan to build an AI data center in Memphis, Tennessee. This deal between the two companies could have cost around $10 billion.

It is buying new Nvidia chip architecture developed by Dell and Supermicro to install in its data center. Based on previous business, xAI has around 16,000 Nvidia chips purchased through Oracle.

Reacting to this report, xAI founder, Elon Musk said his company has contracted for 24,000 Nvidia H100 chips from Oracle. These will be used to train the Grok 2 large language model. It is also revealed that xAI is building a new data center with more than 100,000 H100 systems for fast processing. Training on this system is likely to begin later this month.

“It will be the most power training cluster in the world by a large margin” wrote Musk.

He explained that building 100,000 H100 systems internally to achieve competitiveness and fast processing power in the AI industry.

“Oracle is a great company and there is another company that shows promise also involved in that OpenAI GB200 cluster, but, when our fate depends on being the fastest by far, we must have our own hands on the steering wheel, rather than be a backseat driver” Musk concluded.

xAI team and supercomputer infrastructure (Source – xAI)

The server rack setup is still underway at the xAI facility and the company shared a photograph last week to update on the progress. Half of these racks are provided by Dell and the rest are built in-house. Supermicro computer is providing a cooling solution to achieve better cooling for the server infrastructure.

It’s expected that the first major cluster will go online within next month along with Grok 2 release.