AMD launches 6th Gen EPYC server CPUs for AI data centres
AMD's Zen 6-based server processors are now shipping, with cloud deployments and AI infrastructure support on the way.
AMD has announced general availability of its 6th Gen EPYC SP7 server processors, bringing its Zen 6 architecture to systems now shipping from leading server manufacturers. Microsoft Azure and Oracle Cloud Infrastructure are expected to make matching cloud instances generally available soon, according to AMD.1
The launch is aimed squarely at the data centre rather than consumer PCs. AMD says the new EPYC range is designed for enterprise services, cloud computing, high-performance computing, GPU host nodes and agent sandboxes — the CPU-heavy parts of modern AI infrastructure that sit around the accelerators.
What AMD is launching
The portfolio includes processors tuned for different jobs instead of one all-purpose configuration. AMD lists high-frequency host-node models reaching up to 5.3GHz, alongside the EPYC 9996 with 256 cores. The platform also supports up to 96 lanes of PCIe Gen 6 connectivity, giving server builders more bandwidth for accelerators, storage and networking.1
AMD is also pushing the chips as a way to run more AI agents within a fixed power budget. Its own tests and estimates claim performance and efficiency gains against rival server platforms, including up to 3.7 times the performance of an Nvidia Vera CPU-based system for one EPYC 9996 configuration. Those comparisons come from AMD's testing and modelling, however, so real-world results will depend on memory, software, cooling, orchestration and the workload in question.1
The company says systems are already shipping from leading server vendors, while Azure has preview instances across AI data processing, agentic AI and high-performance computing. Oracle's E7 cloud instances are also planned to use 6th Gen EPYC processors. Independent reporting from SiliconANGLE confirms the same-day Zen 6 EPYC launch, but does not independently validate AMD's comparative benchmark claims.2
Why it matters
The important part of this launch is availability, not just the specification sheet. AMD is placing EPYC 6th Gen into the server supply chain now, giving cloud providers and enterprise customers a new x86 option as AI workloads make CPUs responsible for more scheduling, data handling and agent execution. Pricing and consumer availability were not announced, and most buyers will encounter these processors through a server vendor or cloud instance rather than as a retail chip.
Our opinion
AMD's most credible pitch is that AI infrastructure needs a portfolio of CPUs rather than a single headline part. The company still has to prove its efficiency and agent-capacity claims outside its own test conditions, but shipping hardware and named cloud deployments give this launch more weight than another future-facing AI announcement. The next useful evidence will be independent benchmarks and actual cloud pricing.
1 AMD's figures and availability claims are attributed to its announcement and should be read alongside its stated testing caveats. 2 SiliconANGLE's report independently confirms the launch timing and Zen 6 server focus. 3 AMD's product page provides the official EPYC 9006 platform details and product imagery.