xAI, Elon Musk’s AI startup, is planning to build a massive supercomputer to enhance its AI chatbot, as reported by Reuters quoting Information. This project, referred to as the Computing Gigafactory, is expected to be completed by the fall of 2025 and may involve a collaboration with Oracle. The supercomputer aims to surpass rival GPU groups in both size and capacity.
Elon Musk revealed in an investor presentation that the new supercomputer will utilize up to 100,000 Nvidia H100 GPUs based on the Hopper architecture, which would make it at least four times larger than the largest existing GPU clusters. Despite the high demand for Nvidia’s H100 GPUs in the AI data center chip market, xAI is opting for this previous-generation technology for its 2025 supercomputer. This hardware investment underscores the scale of xAI’s ambitions, with Musk personally taking responsibility for delivering the project on time.
xAI is positioning itself as a direct competitor to AI giants like Open AI and Google, with Musk emphasizing the importance of this supercomputer in developing large language models. The Grok 2 model required around 20,000 Nvidia H100 GPUs for training, and future iterations, such as Grok 3, could necessitate up to 100,000 GPUs.
While xAI and Oracle have not commented on their collaboration, Musk’s dedication to advancing artificial intelligence technology through major investments in infrastructure and strategic partnerships is clear. The exact details of the partnership and the supercomputer project remain open to speculation due to the lack of official statements.
In conclusion, xAI’s plan to build a massive supercomputer with Nvidia H100 GPUs represents a significant step towards enhancing its AI capabilities and competing with industry leaders in the AI space. With Elon Musk leading the charge, xAI is poised to make a mark in the realm of artificial intelligence technology, setting the stage for future innovations in the field.
Article Source
https://www.tomshardware.com/tech-industry/artificial-intelligence/elon-musks-xai-plans-to-build-gigafactory-of-compute-by-fall-2025-using-100000-nvidias-h100-gpus