- Bloomberg stories the contact is for Elon Musk’s X social media platform
- xAi’s Colossus makes use of 100,000 Nvidia H100 GPU and does not appear to be a part of the contract
- Supermicro has had troubles, however continues to be a serious participant in AI {hardware}
The rising demand for computing energy to help AI workloads has fueled fast development out there for high-powered servers. It’s a profitable enterprise for corporations like Dell, Supermicro, and Hewlett Packard Enterprise, who’ve all seen elevated demand for his or her high-performance server merchandise lately.
A report from Bloomberg claims HPE has secured an enormous deal value greater than $1 billion to provide Elon Musk’s X social community with servers optimized for synthetic intelligence work.
The report doesn’t specify precisely how the servers might be used, however because it’s for X, there’s a very good likelihood a number of the capability might be for Grok, the social community’s AI chatbot. On the finish of 2024, X introduced that it was rolling out Grok-2 to all customers without spending a dime, and it’s honest to conclude that additional capability would require extra server infrastructure.
Dell and Supermicro outbid
Bloomberg says X’s settlement with HPE was reached in late 2024, in line with folks accustomed to the matter.
Musk’s firms, together with xAI and Tesla, are main patrons of AI {hardware}. The Colossus supercomputer, constructed by xAI in Memphis, was beforehand declared by Musk to be the “strongest AI coaching system on the earth.” The billionaire said it was constructed “from begin to end” in simply 122 days.
Colossus makes use of 100,000 Nvidia H100 GPUs, however there are plans to massively increase its operations. Supermicro has confronted some challenges just lately, together with the resignation of its auditor and different associated issues. Nonetheless, as MarketWatch stories, it stays a “main participant within the AI revolution” and just lately established operations in Memphis to help xAI’s purpose of constructing a supercomputer facility with 1 million GPUs.
Supermicro and Dell supplied the servers for Colossus initially, and Bloomberg says that whereas each corporations did bid to provide the tools for this new enterprise they had been finally unsuccessful.
Bloomberg’s report notes, “HPE’s liquid-cooling know-how might have performed a task within the win,” wrote Woo Jin Ho, an analyst at Bloomberg Intelligence. Whereas good for gross sales, bigger offers is usually a drag on margins, he added.”
