Chipmaker AMD is taking goal at competitor Nvidia with its newest {hardware} launch: a rack-scale system designed to energy computing wants of the world’s largest AI labs.
On the firm’s sold-out Advancing AI convention in San Francisco on Thursday, AMD Chair and CEO Dr. Lisa Su promoted the brand new AI rack system often known as Helios — together with its rising checklist of consumers, together with Microsoft — as the corporate prepares to ship it later this 12 months. Su additionally pitched the corporate’s latest chips which can be designed to feed the compute-hungry dragon that’s the AI trade.
Rack methods mix many processors right into a single high-powered unit. They’re constructed for knowledge facilities, the place they practice and run AI fashions and different compute-intensive workloads.
Su known as Helios the tech trade’s “highest efficiency AI rack,” including that it was “constructed to coach and run probably the most demanding frontier fashions on the earth at huge scale.” The system shall be deployed by main AI firms at gigawatt-scale, the corporate stated.
Nvidia has traditionally dominated this market with its Vera Rubin and Grace Blackwell rack-scale methods. AMD is clearly seeking to get in on the motion. And Helios’ efficiency metrics seem to provide it an actual probability, beating out Vera Rubin by a lot of metrics, the Register reported.
Helios, which was first revealed in 2025 and shown onstage in January at CES 2026, already has a number of well-known prospects, together with OpenAI, Meta, Oracle, Anthropic, and Microsoft, all of which have plans to deploy the system. Microsoft CEO Satya Nadella said Monday that the corporate would develop its Azure infrastructure with Helios. In the meantime, Anthropic and AMD announced a strategic partnership Wednesday for deploy as much as two gigawatts of GPUs through the brand new rack system.
AMD also introduced Thursday its Venice-X CPU, which is designed for knowledge facilities and to deal with high-computing workloads. The Venice-X is anticipated to launch in 2027.
Throughout her remarks, Su commented on the trajectory of the chip trade, claiming that, by the 12 months 2030, chips that energy AI will turn into an enormous a part of the general computing market. This’s as a result of the trade is “seeing a step change in compute demand” pushed largely by the rise of agentic AI, she stated.
“Once you ask the agent to do one thing, it really has dozens of steps, and it has to purpose, and it has to name instruments, and it has to entry knowledge, and it has to maintain doing it time and again till it solves the issue, and so that you want a number of GPUs to do all that,” the manager stated.
“We’re now anticipating that by 2030, the AI accelerator market goes to achieve about $1.4 trillion,” Su stated. “What which means is, by the tip of the last decade, the AI accelerator market goes to strategy the dimensions of the complete semiconductor market at present.”
“We do count on that GPUs are going to make up the overwhelming majority of that market as a result of the algorithms are nonetheless very a lot of their infancy, and we’re nonetheless persevering with to see the workloads change, and that favors programmability within the total silicon ecosystem,” she added.
Once you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.
