Elon Musk’s Grok AI ‘compute factory’ will use Dell and Supermicro servers

The supercomputer facility that will power Elon Musk’s new artificial intelligence (AI) chatbot, Grok, will be built as part of a hardware collaboration with Dell and Super Micro Computer (Supermicro), it has been announced.

News of the companies’ involvement in the new datacenter that will house the would-be ChatGPT rival emerged on the X (Twitter) platform from Dell CEO, Michael Dell, who tweeted: “We’re building a Dell AI factory with Nvidia to power Grok for xAI.”

In fact, this will be a joint enterprise, Musk tweeted back shortly after, “To be precise, Dell is assembling half of the racks that are going into the supercomputer that xAI is building.”

From other announcements, we know that Dell and Supermicro will build the Grok supercomputer cluster using Nvidia’s latest Blackwell GPU platform, announced in March.

The image accompanying Dell’s tweet showed lines of rack-bound Nvidia servers still in their plastic-wrapped shipping state, which underlines that the project to build the dedicated facility in an unconfirmed location is still in its early stages.

AI gigafactory

Currently, for its development phase, xAI’s chatbot uses the Grok 1.5/1.5v multimodal large language model (LLM) whose big feature is that it can process images, speech, and video in addition to text.

However, as of today, Grok is only available to X’s $16 per month Premium+ subscribers. Presumably that’s because the physical capacity of the system remains limited. Musk has previously said he wants to have Grok to reach its higher-capacity supercomputer incarnation — housed in what Musk has dubbed “the gigafactory of compute” — by late 2025.

The whole project is a cocktail of old and new ingredients of the sort the tech industry excels at. Elon Musk is the most famous tech entrepreneur of his generation. The startup he founded to build Grok, xAI, recently raised a huge $6 billion in series B funding.

Nvidia, whose hardware will be used by Dell and Supermicro, now vies with Microsoft to be the world’s most valuable company with a market cap around $3 trillion.

Meanwhile, the world can’t get enough of a new generation of chatbots, a sector Musk thinks Grok can dominate. Dell and Supermicro are old hands by comparison but have a long track record of making computer systems.

The term “Grok” is jargon used loosely in programming to signify that something is fully intuited or understood. Its origin is widely claimed online to be from Robert A. Heinlein’s 1961 science fiction novel, Stranger in a Strange Land.

If Musk and Nvidia’s GPUs get most of the publicity around Grok, the involvement of Dell and Supermicro is at least as significant. Supermicro, for example, is known for its cooling expertise, a critical part of any datacenter clustering, as well as its understanding of the Nvidia platform.

Key differentiators

A bigger issue for the Grok project might simply be who it’s for in what is starting to become a crowded and competitive chatbot/LLM field:

“I can’t see what the userbase is for Grok when it’s charging for access and not offering anything to differentiate itself except a, Musk himself and b, its access to Twitter data,” tech commentator Kate Bevan told Network World.