AMD's 'Helios' AI Server Enters Full Production, Securing OpenAI and Anthropic Support

Advanced Micro Devices (AMD) announced its second-generation artificial intelligence server, designed to compete directly with Nvidia's flagship products, is now in full production. CEO Lisa Su confirmed during a keynote address on Thursday that shipments of the new 'Helios' server racks are scheduled to begin at the end of the third quarter.

Key Details and Customer Adoption

The Helios servers are powered by AMD's latest MI455X AI accelerator and its new Venice central processor (CPU), both manufactured by Taiwan Semiconductor Manufacturing Co. Su stated that "customer demand for Helios is extremely strong," a claim supported by endorsements from leading AI companies.

During the event, executives from major AI labs joined Su on stage to announce their plans to adopt the new hardware:

• OpenAI: Sachin Katti, vice president of compute strategy at the ChatGPT creator, said the company will begin deploying Helios racks at "massive scale starting towards the end of this year, and then accelerating throughout 2027." Katti also confirmed OpenAI's intent to use AMD's next-generation MI500 AI chips.
• Anthropic: Co-founder Tom Brown highlighted the platform's ease of use, noting that the company's Claude AI was able to set up AMD's servers independently over a weekend. AMD recently announced a deal to sell its MI450 chips to Anthropic beginning in 2027.
• Cerebras Systems: CEO Andrew Feldman announced a partnership to offer a combination of AMD and Cerebras AI servers, initially within Cerebras's own data centers.

Targeting a $2 Trillion Market

Ad

AMD's push comes as it aims to capture a significant share of the rapidly expanding data center market from its rival, Nvidia, particularly in the field of AI inference. This is the computational process that occurs when a user interacts with an AI model like a chatbot.

Su presented an ambitious market forecast, estimating the total computing market will grow from an estimated $365 billion in 2025 to $2 trillion by 2030. Within that 2030 forecast, AMD projects that AI accelerators will account for $1.4 trillion and CPUs will represent $220 billion.

Market Reaction

Following the announcements, shares of AMD fell approximately 4%, a move an Emarketer analyst attributed to a broader selloff in semiconductor stocks and the U.S. market. In contrast, shares of new partner Cerebras Systems jumped nearly 10%.

The news from AMD follows recent technical disclosures from Nvidia about its upcoming Vera CPU and Rubin graphics processing unit (GPU), underscoring the intensifying competition to supply the hardware powering the AI industry.