Elon Musk plans to build an AI supercomputer with Nvidia graphics cards for his AI company xAI, according to a report from The Information, which cites a presentation Musk gave to investors.
The computer, dubbed the "Gigafactory of Compute," is scheduled to be completed by the fall of 2025 and is projected to be at least four times larger than the most powerful clusters currently used by competitors such as Meta.
Musk personally guarantees the timely completion of the supercomputer, which will use Nvidia H100 GPUs. It's not yet clear where the cluster will be built.
By comparison, Meta aims to have 340,000 Nvidia H100 GPUs and about 600,000 graphics cards in use by the end of the year, Meta CEO Mark Zuckerberg said in January 2024.
Musk's AI startup, xAI, may be partnering with Oracle, which already supports the training and inference of its current AI chatbot, Grok, on Musk's "X" platform. According to The Information, xAI has an Oracle cluster of 16,000 H100 chips for Grok, which is trained on a total of 20,000 GPUs. Nvidia has also reportedly given xAI priority status for shipping its new "Blackwell" AI GPU.
Musk's AI company, xAI, reportedly raised $6 billion in funding at the end of April at a pre-money valuation of $18 billion. The funding round is expected to close in the coming weeks.
In early April, Musk announced a new Grok model coming in May that he claims is better than the GPT-4. He also believes that AI capabilities will surpass human intelligence by the end of next year, given sufficient computing power and energy.
Microsoft and OpenAI are also reportedly planning to build a new supercomputer called "Stargate," which could cost up to $100 billion and be fully developed by 2030. A smaller supercomputer, known as "Phase 4," could be operational as early as 2026. The full expansion of the project depends on whether OpenAI can make significant progress in its AI research toward superintelligence.
