AMD Launches Ambitious AI Server to Rival Nvidia as OpenAI Chooses Its Next-Gen Chips

On Thursday, Advanced Micro Devices CEO Lisa Su introduced a new artificial intelligence server expected for 2026, designed to rival Nvidia’s top products. Meanwhile, OpenAI’s CEO announced that the creator of ChatGPT would utilize the newest AMD chips.

Su addressed attendees at a developer conference named «Advancing AI» in San Jose, California, to showcase the MI350 and MI400 series AI chips, which she claims will go up against Nvidia’s Blackwell processor lineup.

The MI400 chip family will lay the foundation for a new server known as «Helios,» which AMD anticipates launching next year.

This strategic shift underscores the evolving competition between Nvidia and other AI chip manufacturers, focusing increasingly on comprehensive server solutions featuring numerous processors and integrated networking chips from the same vendor.

According to AMD executives, the Helios servers will incorporate 72 MI400 chips, positioning them against Nvidia’s existing NVL72 servers.

During the keynote, AMD indicated that many components of the Helios server, including networking protocols, would be shared openly with competitors like Intel.

This announcement is a clear challenge to Nvidia, which relies on its proprietary NVLink technology to connect its chips, although it has recently begun licensing this technology amid increasing pressure from competitors.

«The future of AI will not be dictated by one company or a closed ecosystem. It will thrive through open industry collaboration,» stated Su.

On stage with Su was OpenAI’s Sam Altman, whose organization is collaborating with AMD on refining the MI450 chips for enhanced AI applications.

«The growth of our infrastructure over the past year and our vision for the coming year have been incredibly exciting,» said Altman.

During her address, representatives from Elon Musk-owned xAI, Meta Platforms, and Oracle shared insights on how they utilize AMD processors. Crusoe, a cloud service specialized in AI, revealed plans to purchase $400 million (approximately Rs. 3,440 crore) worth of AMD’s upcoming chips.

Su reiterated the company’s roadmap for the upcoming year, aligning it closely with the annual release strategy first established by Nvidia with its Blackwell chips.

Following the announcement, AMD’s stock fell by 2.2 percent. Analyst Kinngai Chan of Summit Insights noted that the chips revealed on Thursday are unlikely to immediately enhance AMD’s competitive standing.

Although AMD has encountered challenges in capturing market share in the burgeoning AI chip arena dominated by Nvidia, the company has actively sought to enhance its software offerings and create a competitive chip lineup.

In March, AMD finalized the acquisition of server manufacturer ZT Systems, leading to expectations of new AI systems akin to some of Nvidia’s server-rack-size products.

AMD, headquartered in Santa Clara, California, has pursued a series of minor acquisitions in recent weeks and expanded its talent pool in chip design and AI software. Su disclosed that the company made 25 strategic investments last year related to its AI initiatives.

Last week, AMD welcomed the team from the chip startup Untether AI, and on Wednesday, it announced the hiring of several key personnel from generative AI startup Lamini, including the co-founder and CEO.

AMD’s ROCm software has struggled to compete against Nvidia’s CUDA, considered by many insiders to be a critical element in maintaining Nvidia’s market dominance.

When AMD released its earnings report in May, Su remarked that despite rising restrictions on AI chip exports to China, the company remained optimistic about achieving robust double-digit growth from AI chips.

© Thomson Reuters 2025

(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)