U.S. chipmaker AMD has decided to deploy its artificial intelligence (AI) infrastructure system "Helios" on Microsoft's cloud platform "Azure." Analysts say AMD is shaking Nvidia's dominance by winning supply contracts from major AI companies one after another, following Meta (formerly Facebook) and OpenAI.
On the 20th, AMD announced it would supply its AI rack system "Helios" to Microsoft Azure. An AI rack is a large server system that integrates graphics processing units (GPUs), central processing units (CPUs), networking, and software into a single cabinet.
Through Helios, AMD is entering the rack AI system market in earnest, moving beyond chips, and will now compete with Nvidia's flagship AI systems, Grace Blackwell and Vera Rubin. AMD plans to supply Helios to major customers, including Microsoft, starting late this year.
According to market research firm The Futurum Group, Nvidia currently holds about 95% of the data center GPU market, while AMD's share stands at about 4.5%. However, analysts say that AMD, which until now had only supplied standalone chips, will now provide a platform integrating its own network technology and software to a large cloud company such as Microsoft, marking an opportunity to expand its influence in the AI infrastructure market.
"Token value" is also cited as an AMD advantage. Helios itself is estimated to be more expensive than Nvidia's Vera Rubin. The Futurum Group estimated the price of Helios at $5 million to $5.5 million. That is higher than the expected price of $3.5 million to $4 million for Vera Rubin, Nvidia's second-generation rack-scale system.
AMD's customer base is expanding. The company said that eight of the top 10 AI companies use its Instinct GPU, AMD's GPU brand for AI, with customers including OpenAI and SpaceX AI.
Neil Shah, an analyst at Counterpoint Research, said, "Helios's GPU and CPU performance is on par with Nvidia's, but Nvidia is differentiated from AMD in terms of software and optimization." He explained that Nvidia's ecosystem, centered on its parallel computing platform "CUDA," is being utilized in a uniquely dominant way in the industry.
However, Shah predicted that AMD could make a leap forward if early customers achieve good operational results, adding, "The key question will be whether AMD wins because it is technically superior, or whether it succeeds simply because the AI chip supply shortage is so severe that anything made will sell."
