AMD’s Next Server CPUs Are Already Spoken For as AI Agents Reshape Data Centers

However, even before the next generation of AMD server CPUs hits the market, there is a rather unexpected problem. Apparently, all the planned 2027 supply of EPYC 9006 processors has already been sold out. Furthermore, some customers are considering purchasing AMD server CPUs for 2028.

It is not only the server load that has changed. AI agents do more than reply to prompts; they schedule actions, make calls to external services, manipulate data, execute code, etc. The CPU will have to work much harder than in the past.

Until recently, most of the attention in the AI infrastructure was paid to GPUs. However, now CPUs can become the bottleneck, and if the processor is not able to transfer the information and control the processes, the expensive GPU will be just sitting idle.

EPYC 9006 has been developed precisely for such a type of load. It provides PCIe 6.0 interface and the memory bandwidth up to 1.6 TB/s.

AMD Expects a Huge Server CPU Business in 2027

Morgan Stanley projects AMD to be able to ship about 6.75 million EPYC CPUs in the coming year. At an estimated ASP of $7,691, the revenue generated from such shipment would total around $51 billion. Prices for server CPUs have increased by over 40% amid supply issues.

The expansion of the EPYC 9006 series is done in stages, with the SP7 version being able to accommodate up to 256 cores, followed by the Venice-X and Verano versions.

Our Take

This change is not only relevant to AMD and data center companies. With more integration of AI agents into regular software, the supporting hardware will require increased processing capabilities. While you might not be aware of the CPU that an AI service runs on, its presence will affect capacity, prices and finally the price of the product.

Want to keep up with major AI and hardware developments? Share this article and follow us on social media. You can also explore our other technology stories for more insights.