We previously revealed how Microsoft and OpenAI are rumoured to be collaborating on a ground-breaking data centre project called 'Stargate', set to launch in 2028. Funded by Microsoft to the tune of over $100 billion, the project aims to reduce the company's reliance on Nvidia – something many tech giants involved in AI are looking to do these days.
In April, The next platform He said this AI supercomputer would likely be “based on future generations of Cobalt Arm server processors and Maia XPUs, with Ethernet scalability from hundreds of thousands to 1 million XPUs on a single machine.”
While we don't yet know much about the actual details, or even if Stargate will ever happen (it's certainly an ambitious project), Forrest Norrod, AMD executive vice president and general manager of the Data Center Solutions Group, made an intriguing claim in a conversation with The next platform.
Without naming names
Speaking of a future Instinct MI500 GPU accelerator, NTP (National Transition Plan)Timothy Prickett Morgan asked Norrod, “What's the largest AI training cluster that anyone is seriously considering? I don't need to name names. Has anyone come to you and said that with MI500 I need 1.2 million GPUs or something?”
Norrod replied: “Is it in that range? Yes.” Asked for more details, he added: “I’m serious, it’s in that range,” and then clarified: “I’m talking about a machine… The scale of what’s being contemplated is mind-boggling. Now, will all of that become a reality? I don’t know. But there are public reports of very sensible people who are contemplating spending tens of billions of dollars or even a hundred billion dollars on training clusters.”
Few companies could afford such a “mind-boggling” project with over a million GPUs, so it's not far-fetched to link any conversations AMD has had on the subject to Stargate. Turning to AMD would certainly make a lot of sense for any company looking to bypass Nvidia…
