Could AMD Hold the Key to Microsoft’s Amazing $100 Billion Stargate AI Supercomputer? AMD EVP Reveals Plans for Million+ GPU Training Cluster Based on Future MI500 Chip — But Keeps Quiet on Customer
We previously revealed that Microsoft and OpenAI are reportedly collaborating on a groundbreaking data center project called ‘Stargate’, set to launch in 2028. The project, which is being funded by Microsoft to the tune of over $100 billion, aims to reduce the companies’ reliance on Nvidia – something that many tech giants involved in AI are currently looking for.
In April, The next platform said this AI supercomputer would likely be “based on future generations of Cobalt Arm server processors and Maia XPUs, with Ethernet that can scale from hundreds of thousands to 1 million XPUs in a single machine.”
While we still don’t know much about the actual details, or even whether Stargate will ever happen (it’s certainly an ambitious undertaking), Forrest Norrod, AMD’s EVP and general manager of the Datacenter Solutions Group, made an intriguing claim in an interview with The next platform.
Without mentioning names
When we talk about a future Instinct MI500 GPU accelerator, TNPTimothy Prickett Morgan asked Norrod, “What’s the largest AI training cluster that anybody is serious about? You don’t have to name names. Has anybody come to you and said with MI500, I need 1.2 million GPUs or something like that.”
Norrod responded: “It’s in that range? Yes.” When pressed for more detail, he added: “I mean it, it’s in that range,” and then clarified: “I’m talking about one machine… The scale of what’s being considered is staggering. Is it going to happen? I don’t know. But there are public reports from very sober people who are considering spending tens of billions of dollars or even hundreds of billions of dollars on training clusters.”
Few companies could afford such a mind-blowing project involving over a million GPUs, so it’s not unreasonable to link discussions AMD has had on the matter to Stargate. Turning to AMD would certainly make a lot of sense for any company looking to bypass Nvidia…