AMD CEO Lisa Su introduced a major new line of AI hardware on Thursday, unveiling both the MI350 and MI400 series of AI chips and announcing plans to release the company’s first AI server, called “Helios,” in 2026. The launch signals AMD’s most direct challenge yet to Nvidia’s dominance in the AI server and chip market.
The announcement was made at AMD’s “Advancing AI” developer conference in San Jose, California. The Helios servers will house 72 MI400 chips, designed to compete directly with Nvidia’s current NVL72 servers powered by its Blackwell processors. In a notable difference from Nvidia’s closed ecosystem, Su emphasized that many aspects of AMD’s server and networking standards would be openly available to the broader industry, including rivals like Intel.
“The future of AI is not going to be built by any one company or in a closed ecosystem. It’s going to be shaped by open collaboration across the industry,” Su stated.
Su was joined on stage by OpenAI CEO Sam Altman, who confirmed that OpenAI is already working with AMD on its MI450 chips to help optimize them for AI workloads. Altman remarked on OpenAI’s rapid infrastructure growth, calling the pace “crazy” and noting continued expansion with AMD’s hardware.
Executives from Meta Platforms, Elon Musk’s xAI, and Oracle also appeared during the event to showcase how their companies are adopting AMD’s processors. Additionally, Crusoe, a cloud provider specializing in AI, disclosed plans to purchase $400 million worth of AMD’s new chips.
Despite the announcement, AMD shares slipped 2.2%, with analysts suggesting the company still faces significant headwinds in dislodging Nvidia’s dominant market position. Summit Insights analyst Kinngai Chan noted that the newly announced products are unlikely to shift the competitive balance immediately.
AMD has aggressively expanded its AI capabilities over the past year, completing its acquisition of server manufacturer ZT Systems in March and making 25 strategic investments in AI-related startups. The company recently hired engineers from Untether AI and Lamini, further strengthening its chip design and software development teams.
However, AMD’s ROCm software stack continues to lag behind Nvidia’s highly entrenched CUDA platform, which many in the industry see as a major factor behind Nvidia’s dominance.
Nevertheless, AMD remains optimistic about its growth prospects, even as U.S. export controls tighten on AI chip sales to China. When reporting earnings in May, Su reiterated expectations for strong double-digit growth in AI chip sales despite these headwinds.