Chips & Hardware
AMD unveils Helios AI rack system to challenge Nvidia
AMD showed off its Helios AI rack-scale system at its Advancing AI conference, naming Microsoft, OpenAI, Meta, Oracle, and Anthropic as customers ahead of a shipment later this year.
AMD used its sold-out Advancing AI conference in San Francisco on Thursday to unveil Helios, a rack-scale AI system the company is preparing to ship later this year. AMD Chair and CEO Lisa Su promoted the system alongside a growing customer list that includes Microsoft, and pitched AMD’s newest chips as a response to surging compute demand across the AI industry. Rack-scale systems combine many processors into a single high-powered unit built for data centers, where they train and run AI models and other compute-intensive workloads.
Nvidia has historically dominated this market with its Vera Rubin and Grace Blackwell rack-scale systems, and AMD is positioning Helios to compete directly. Helios beats Vera Rubin by a number of performance metrics, The Register reported. The system was first revealed in 2025 and shown onstage at CES 2026 in January; it already counts OpenAI, Meta, Oracle, Anthropic, and Microsoft among customers with plans to deploy it. Microsoft CEO Satya Nadella said Monday that the company would expand its Azure infrastructure with Helios, and Anthropic and AMD announced a strategic partnership Wednesday to deploy up to two gigawatts of GPUs via the new rack system. AMD also introduced its Venice-X CPU on Thursday, built for data centers and high-computing workloads and expected to launch in 2027.
Su framed the launch against a broader shift in the chip industry, describing a step change in compute demand driven largely by the rise of agentic AI — systems that reason through dozens of steps, call tools, and access data repeatedly to solve a problem, with each step requiring more GPU capacity. “We’re now expecting that by 2030, the AI accelerator market is going to reach about $1.4 trillion,” Su said, adding that by the end of the decade the AI accelerator market will approach the size of today’s entire semiconductor market. She said GPUs should keep making up the vast majority of that market because AI algorithms and workloads are still evolving, which favors programmable silicon over more specialized designs.
Why it matters
AMD’s push at gigawatt scale — with Anthropic alone committing to up to two gigawatts of GPUs — signals that the AI rack-scale market Nvidia has long dominated is becoming genuinely contested, with implications for how AI labs price and source the compute behind frontier models.