On Thursday, Lisa Su, CEO of Advanced Micro Devices (AMD), introduced a new artificial intelligence server set for release in 2026, directed at competing with Nvidia’s leading products. Her announcement coincided with remarks from OpenAI’s CEO, who confirmed that the ChatGPT developer plans to utilize AMD’s latest chip technology.
During the “Advancing AI” developer conference in San Jose, California, Su outlined the capabilities of the MI350 and MI400 series AI chips designed to rival Nvidia’s Blackwell processor line.
The MI400 series will serve as the foundation for a forthcoming server named “Helios,” which AMD intends to launch next year.
This development reflects a shift in the competition landscape among AI chip manufacturers, moving from proprietary chip sales to the provision of full servers that integrate multiple processors and networking components produced by the same company.
The new Helios servers will incorporate 72 MI400 series chips, positioning them to compete with Nvidia’s NVL72 servers, according to AMD officials.
In her keynote address, Su emphasized that many features of the Helios servers, including networking standards, would be made publicly accessible and shared with competitors like Intel.
This approach directly targets Nvidia, known for its proprietary NVLink technology used to connect its chips. Recently, however, Nvidia has begun licensing that technology amidst rising competition.
“The future of AI will not be defined by a single entity or a closed ecosystem, but rather through collaborative efforts across the industry,” Su stated.
OpenAI’s Sam Altman joined Su on stage, discussing collaboration with AMD to enhance the design of their MI450 chips for AI applications. “The infrastructure ramp-up over the past year has been astonishing to witness,” Altman remarked.
Executives from multiple companies, including Elon Musk-owned xAI, Meta Platforms, and Oracle, also presented their applications of AMD technology during the event. Crusoe, a cloud provider focused on AI, announced plans to procure $400 million worth of AMD’s new chips.
Su reaffirmed the company’s roadmap for the upcoming year, aligning closely with the annual product cadence established by Nvidia’s Blackwell chips.
Despite the optimistic announcements, AMD’s stock dipped by 2.2 percent following the presentation. Analyst Kinngai Chan from Summit Insights noted that the newly revealed chips are unlikely to significantly alter AMD’s competitive landscape in the near term.
AMD has encountered challenges in gaining market share within the rapidly expanding AI chip sector dominated by Nvidia. However, the company has been actively improving its software offerings and developing chip lines that can compete on performance.
This March, AMD completed its acquisition of ZT Systems, a server builder, which is expected to facilitate the launch of fully integrated AI systems akin to Nvidia’s server products.
Based in Santa Clara, California, AMD has undertaken several small acquisitions in recent weeks and bolstered its chip design and AI software teams. Su mentioned that the company has made 25 strategic investments over the past year focused on enhancing its AI initiatives.
Last week, AMD recruited the team from chip startup Untether AI, and on Wednesday, announced the hiring of several employees from generative AI startup Lamini, including its co-founder and CEO.
Despite challenges with its ROCm software gaining traction against Nvidia’s CUDA, which many industry insiders view as a critical barrier to Nvidia’s market dominance, AMD continues to anticipate robust double-digit growth from its AI chip sector, even in light of increasing restrictions on AI chip exports to China, as reported during their earnings call in May.
© Thomson Reuters 2025
(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)
