NEW YORK, May 27: In a recent presentation to investors, US entrepreneur Elon Musk revealed plans by his artificial intelligence startup xAI to develop a supercomputer to power the next iteration of its AI chatbot Grok, as reported by The Information on Saturday.

Musk aims to have the proposed supercomputer operational by fall 2025, with discussions underway for a potential partnership with technology giant Oracle for its development.

While xAI could not be reached immediately for comment, Oracle did not respond to requests for comment from Reuters.

The envisioned supercomputer will rely on connected groups of Nvidia’s flagship H100 graphics processing units (GPUs), estimated to be at least four times larger than current GPU clusters. Musk presented these details to investors in a May presentation cited by The Information.

Nvidia’s H100 GPUs are highly sought after in the data center chip market for their powerful AI capabilities, but their availability can be limited due to high demand.

Founded by Musk last year, xAI aims to rival established players like Microsoft-backed OpenAI and Alphabet’s Google in the AI space. Musk is also a co-founder of OpenAI.

Earlier this year, Musk disclosed that training the Grok 2 model required approximately 20,000 Nvidia H100 GPUs, while future iterations like Grok 3 will demand even more computational power, necessitating around 100,000 Nvidia H100 chips.