AMD EPYC™ 9006 SP7-based Platforms Are Shipping: Advance Your Infrastructure for Agentic AI

AMD EPYC™ 9006 SP7-based platforms are shipping now, giving customers access to a processor family built for the full range of compute demands reshaping modern infrastructure, from agentic AI to high-performance computing (HPC) and virtualization. AMD EPYC 9006 SP7 Server CPUs are the first of four planned processor families in the portfolio, with AMD EPYC 9006 SP8, 9006X SP7, and 9006 LP Server CPUs to follow, giving customers purpose-built options to scale agentic AI and demanding workloads. For IT and infrastructure leaders planning next-generation deployments, that means the CPUs needed to run agents, keep GPUs productive, and support enterprise workloads are available today.

The Best CPU for the Agentic AI Era¹

AMD EPYC 9006 SP7 Server CPUs lead the AMD EPYC 9006 Series portfolio in core count and performance. Built to serve the Agent Sandbox, AI Host Node, and General-Purpose CPU roles, AMD EPYC 9006 SP7 Server CPUs give customers the scale to run more agents, keep AI accelerators productive, and handle demanding general-purpose workloads while achieving exceptional total cost of ownership. The family brings these first-to-market capabilities:

• Highest number of cores/threads combination with up to 256 cores and 512 threads²
• Frequencies up to 5.3 GHz with 96 cores³
• Support for JEDEC-standard MRDIMM-12800 delivering up to 1.6 TB/s of memory bandwidth per socket⁴
• PCIe® Gen 6 I/O for x86 platforms⁵

Scale Agent Execution

Agentic workflows don't run as a single process. They fan out into many concurrent agents, tool calls, and subagent tasks, and that concurrency needs dense, efficient compute to run on. AMD EPYC 9006 SP7 Server CPUs provide the core and thread density these workflows demand: up to 256 cores and 512 threads per socket let customers run more agents and the services they call within the same rack footprint, while leadership performance/watt helps maximize agent capacity within a fixed power envelope.

Keep AI Accelerators Productive

Agentic AI depends heavily on GPUs for inference, but it's the CPU that keeps those GPUs supplied with data and work. AMD EPYC 9006 SP7 Server CPUs deliver the highest frequency in its class, up to 5.3GHz, for leadership per-core performance in host-node roles⁶, paired with a PCIe Gen 6 I/O subsystem that provides twice the bandwidth per lane versus the previous generation. Combined with up to 1.6TB/s of memory bandwidth, AMD EPYC 9006 SP7 Server CPUs help move large context windows and datasets fast enough to keep accelerators productive rather than waiting on data. 

Handle Demanding Workloads Beyond AI

Agentic systems don't run in isolation. To get work done, agents use tools and connectors to query databases, call APIs, search enterprise knowledge, and access business applications. These activities rely on the same general-purpose CPU infrastructure enterprises already use, with demand growing as more agents are deployed. AMD EPYC 9006 SP7 Server CPUs bring the performance and capacity to support these workloads. With up to 512 threads available, IT teams can consolidate virtual machines onto fewer servers. For HPC workloads, AMD EPYC 9006 SP7 Server CPUs help shorten time to insight for complex simulations across life sciences, quantum chemistry, and weather forecasting through high memory bandwidth and leadership compute performance, based on published SPECrate® benchmark results.⁷

Extending Customer Choice with Cloud Offerings

Cloud providers offer AMD EPYC™ Server CPUs through cloud instances and their internal SaaS software, giving customers flexible ways to deploy compute for AI data preparation, agent orchestration, HPC, and enterprise applications. A wide range of cloud instances, including general-purpose, AI-optimized, and HPC, are based on AMD EPYC 9006 SP7 Server CPUs.

Oracle Cloud Infrastructure (OCI) is expanding that choice with upcoming OCI Compute E7 instances powered by AMD EPYC 9006 Server CPUs. These instances will support enterprise applications and the CPU services that underpin agentic AI, including agent orchestration, concurrent agent sandboxes, and retrieval services.

Microsoft is bringing that choice to advanced AI model development on Azure.

Available Now

AMD EPYC 9006 SP7-based platforms are now shipping from leading OEM/ODM partners. In addition, instances from major cloud service providers (CSP) are available. For IT and infrastructure leaders evaluating how to scale agentic AI alongside existing HPC and virtualization demands, AMD EPYC 9006 SP7 Server CPUs deliver the performance and scale to run more agents, get more from GPU accelerators, increase VM density, and accelerate time to insight for complex HPC simulations, all at exceptional total cost of ownership.

Visit the AMD EPYC 9006 SP7 Server CPUs product page to explore how to advance your infrastructure for agentic AI and find available platforms to get started.