AMD Releases Zen 6 EPYC Chips for Denser AI Agent Deployments
What happened
AMD introduced its sixth-generation EPYC SP7 server processors on November 10, 2026. These new chips are available through server manufacturers, with Microsoft Azure and Oracle Cloud Infrastructure planning to offer them in upcoming cloud instances. The processors are built to handle AI agent tasks, GPU hosting, and enterprise workloads, and are included in AMD's Helion rack systems for large-scale deployments. Microsoft also launched a preview of its Azure HDv2 virtual machines using these processors, aimed at improving parallel agent execution.
Key details
- Zen 6 EPYC SP7 processors range from 64 to 256 cores and 128 to 512 threads.
- Top models, EPYC 9996 and 9956, each have 1,024 MB of L3 cache and up to 600W power rating.
- Processors support PCIe Gen 6 connectivity.
- AMD's Helion racks can scale up to 73,728 threads in a single rack configuration.
- Microsoft's Azure HDv2 VMs offer more than 500 physical Zen 6 cores and 3.8 TB DDR5-8800 memory.
Why it matters
Cloud providers, enterprise IT teams, and organizations running large AI workloads can now deploy more AI agents in the same physical space and power budget. This allows for higher throughput and efficiency in data centers, potentially reducing costs or increasing capacity for AI-driven applications. Customers using Microsoft Azure and other cloud services will have access to these capabilities as new virtual machines become available.