Musk Wants a 'GigaComputer' For xAI ChatBot: The Information

Elon Musk is looking to expand the horizons of his companies by creating an AI superbrain.

The Tesla chief told investors earlier this month he plans to purchase 100,000 powerful computer chips so his artificial intelligence startup xAI can build a giant supercomputer.

The supercomputer – “at least four times as powerful than the GPU clusters that exist today” – will power a new smarter version of its AI chatbot Grok, according to a report by The Information on Saturday, which cited a presentation Musk gave to investors recently.

ALSO SEE: Nvidia Chip Prices Take a Hit in Duel With China’s Huawei

Musk said he wants a “gigafactory of compute” powered by 100,000 of Nvidia’s H100 graphic processing units (GPUs) running by the fall of 2025, the report said, adding that xAI could partner with Oracle to develop the massive computer.

When completed, the connected groups of Nvidia’s flagship chip would be at least four times the size of “those built by Meta Platforms to train its AI models”, The Information said, quoting Musk from his presentation.

xAI could not be immediately reached for comment. Oracle did not respond to a Reuters request for comment.

Nvidia’s H100 family of powerful GPUs dominate the data-centre chip market for AI but can be hard to obtain due to high demand because of the boom in generative artificial intelligence.

The energy required for this ‘GigaComputer’ will be considerable given each H100 GPU is said to consumer 700 watts, which is equivalent to an average American household (2.5 people), according to a report by Tom’s Hardware.

Huang lauds Tesla’s ‘revolutionary’ FSD tech

Nvidia CEO Jensen Huang said in an interview with Yahoo Finance on Friday that the infrastructure Musk was building was “incredible” and the work Tesla was doing with its Full Self Driving system was “revolutionary”.

“One of the things that’s really revolutionary about version 12 of Tesla’s Full-Self-Driving [system] is that it’s an end-to-end generative model, and it learns from watching videos – surround video.. It learns about how to drive end-to-end using generative AI .. to predict the path and how to steer, how to understand and how to steer the car. The technology is really revolutionary.”

Meanwhile, Huang said another startup that Nvidia was working with a company called Recursion that had built a supercomputer for understanding proteins and generating molecules – for drug discovery.

Aside from work being done by big groups owned by Elon Musk and Meta’s Mark Zuckerberg, there were companies in “so many areas” exploring the use of AI, he said.

“We now have software and AI models that can understand .. and use almost any language .. and that capability is now being deployed at scale in so many industries.”

Musk founded xAI last year as a challenger to Microsoft-backed OpenAI and Alphabet’s Google. Musk also co-founded OpenAI.

Earlier this year, Musk said training the Grok 2 model took about 20,000 Nvidia H100 GPUs, adding that the Grok 3 model and beyond will require 100,000 Nvidia H100 chips.

Jim Pollard with Reuters

NOTE: The report was expanded on May 26, 2024 to include remarks by Nvidia chief Jensen Huang about the use of AI by Musk’s companies and others in many sectors.