HPE may have bagged $1B order from Elon Musk's X for AI servers
That's Cray cray
Hewlett Packard Enterprise has reportedly secured a contract to supply Elon Musk's X, the site better known as Twitter, with more than $1 billion in AI-accelerating servers.
The agreement, first reported by Bloomberg, cites anonymous sources that claim HPE is the latest to benefit from the Tesla tycoon's artificial intelligence ambitions.
Musk's model-building outfit xAI, which makes its generative Grok model mainly available via the billionaire's X, has a combination of Dell and Supermicro systems in its 100,000 GPU Colossus supercomputer in Memphis, Tennessee.
That system is due to be upgraded with an additional 100,000 GPUs, with plans to eventually expand the system to a million accelerators in the future.
These HPE-supplied systems, as reported, are in our view either going to end up at X to run AI inference and training work, or at xAI doing the same. X and xAI are quite tangled; as we said, xAI provides a Claude/ChatGPT/Gemini-like model called Grok that netizens interact with mainly from X. Musk has previously redirected 12,000 Nvidia H100 GPUs originally bound for Tesla to xAI, arguing the parts would have sat unused otherwise.
Last month, X opened Grok and xAI's new image-gen capabilities to all X-ers beyond the previous limitation to X Premium subscribers. This month, xAI also launched a standalone Grok app for iOS. Broader availability of these services likely will necessitate additional compute resources, especially if future Grok models continue to grow in size.
HPE, for its part, is no stranger to accelerated computing. Its Cray division is responsible for building some of the largest supercomputers in the world, including the El Capitan, Frontier, and Aurora supercomputers, which rank among the three most powerful systems on the publicly known global top-500 list.
More recently, HPE announced a series of new liquid-cooled GPU systems available with your choice of Nvidia, AMD, or Intel GPUs and AI accelerators. Which of the three X has opted for has yet to be disclosed, however, we'll note that most deployments by Musk's enterprises have utilized Nvidia accelerators.
- CoreWeave drops £1bn in UK datacenters – but don't expect the latest Nvidia magic just yet
- Nvidia snaps back at Biden's 'innovation-killing' AI chip export restrictions
- UK unveils plans to mainline AI into the veins of the nation
- Tesla, Musk double down on $56B payday appeal
It's also possible X's decision to source systems from HPE may have been driven by a desire to find alternative suppliers in the wake of Supermicro's accounting drama.
The Register reached out to HPE and X for comment, and had not heard back at the time of publication. ®