AMD

AMD EPYC Powers Agentic AI Growth With Rack-Scale Performance Edge

AMD says its EPYC processors deliver superior rack-scale throughput for agentic AI workloads, enabling enterprises to maximize CPU-driven infrastructure performance, efficiency, and scalability within power constraints.

AMD Rack Performance Edge

Agentic AI is reshaping enterprise infrastructure requirements, placing greater emphasis on CPU performance to support the growing number of services operating behind intelligent systems. According to AMD, production agentic AI environments rely on far more than GPUs, requiring substantial CPU resources for orchestration, databases, web services, APIs, caches, and middleware that manage and coordinate workloads.

AMD argues that evaluating processor performance at the rack level provides a more realistic measure of data center capability than individual chip benchmarks. Since modern data centers are constrained by power, cooling, floor space, and operational considerations, the company believes the key metric is how much useful work can be delivered within a fixed rack power budget.

Using a modeled 100-kilowatt rack configuration built on dual-socket platforms, AMD reported that its EPYC 9965 processor delivers an estimated 2.37 times the rack-level throughput of NVIDIA Vera and approximately 1.6 times the performance of Intel Xeon 6980P. The company also projects that the upcoming EPYC “Venice” processor will extend the performance advantage to 3.30 times compared with the NVIDIA baseline.

The analysis covered a range of enterprise workloads commonly associated with agentic AI deployments, including general-purpose computing, server-side Java applications, web serving, key-value stores, in-memory caching, and relational databases. AMD said its processors consistently outperformed competing platforms across the evaluated workload set.

The company noted that higher core density within a fixed power envelope allows organizations to increase service throughput and concurrency without exceeding rack limitations. For enterprises deploying large-scale agentic AI systems, this translates into greater responsiveness, improved infrastructure utilization, and the ability to support more AI agents within existing data center resources.

AMD believes these advantages make EPYC processors a strong foundation for organizations seeking scalable, production-ready infrastructure capable of meeting the expanding demands of agentic AI applications.