AMD EPYC CPUs and Pensando DPUs to Power New Oracle Cloud Instances

AMD said on October 27, 2026, that Oracle Cloud Infrastructure (OCI) Compute E7 instances will be powered by 6th Gen AMD EPYC server processors and that Oracle Acceleron converged network interface cards incorporate AMD Pensando data processing units (DPUs) for AMD CPUs and GPUs on OCI. AMD shared the updates to its multigeneration collaboration with Oracle at Oracle AI World 2026 in a newsroom announcement.

“AMD EPYC processors and Pensando DPUs power workloads across AI, database, cloud-native and enterprise applications,” the company stated. AMD said scaling AI requires powerful compute and efficient infrastructure services, with CPUs coordinating agents, retrieving data and running enterprise applications while networking technology moves data and offloads infrastructure tasks.

6th Gen EPYC Processors Underpin OCI Compute E7

E7 instances use 6th Gen AMD EPYC processors from the EPYC 9006 Server Processor Series, codenamed Venice, according to an Oracle Cloud Infrastructure blog post published the same day. Oracle stated that 6th Gen EPYC processors provide up to 256 cores and 512 threads per socket, up to 1,024 MB of L3 cache, 16 memory channels with up to 8000 MT/s, and PCIe 6.0 connectivity. Oracle described those figures as significant uplifts over 5th Gen EPYC CPUs and said the processors power enterprise workloads more quickly and efficiently and allow the deployment of more agents per instance and per watt.

Oracle (ORCL ) said E7 gives customers a CPU option for agent orchestration, concurrent agent sandboxes, retrieval services and in-memory caches, as well as web, API and enterprise application tiers. In an agentic application, CPUs assemble context, route actions, retrieve knowledge, query databases, parse documents and execute tools, and Oracle said those stages influence application performance alongside model inference. E7 Acceleron Standard virtual machines will offer flexible compute and memory configurations, and Oracle said the standards-based x86 foundation supports continuity with existing enterprise software stacks.

Helios Rack-Scale Platform and Pensando Networking on OCI

Oracle described AMD Helios as an open, co-designed rack-scale AI platform aligned with the Open Compute Project Open Rack Wide design, integrating 72 AMD Instinct MI455X GPUs across 18 four-GPU compute trays alongside 6th Gen EPYC Venice CPUs, AMD Pensando networking and AMD ROCm software. Oracle stated that a full 72-GPU rack provides up to 31 TB of HBM4 memory and up to 1.7 PB/s of aggregate peak memory bandwidth, and positioned Helios for large-scale inference, training and fine-tuning, including frontier AI and sovereign computing needs.

According to the Oracle post, ROCm software supports widely used AI frameworks and serving tools, including PyTorch, JAX, vLLM and SGLang, while ROCm.AI and tools such as ROCm CLI, AMD Skills and Hyperloom assist with environment setup, troubleshooting and workload optimization. Oracle said Helios combines high-bandwidth communication within each rack with networking that connects racks into larger clusters, and that OCI customers can evaluate configurations around their models, request volumes and response-time requirements as capacity needs expand beyond a single rack.

Oracle and AMD have co-engineered Pensando-based DPUs with Oracle Acceleron SmartNICs to move networking, storage, security and infrastructure functions closer to the data path, preserving host CPU capacity for customer applications. The SmartNICs combine host networking and cloud-control functions in a single architecture while maintaining isolation between customer and cloud-managed functions.

AMD stated in a separate blog post that the Pensando-based Acceleron architecture “can deliver up to 2x improvement in network processing efficiency and increased storage performance,” depending on workload and configuration. The architecture supports OCI Zero Trust Packet Routing, which applies access policies close to where traffic enters and leaves the host, and enables security patching with near-zero downtime.

AMD said Pensando DPUs provide a programmable foundation that lets OCI evolve networking, storage, security and observability services as workloads change. The company said training often moves large volumes of data among a relatively fixed set of GPUs in flows that can be long-lived, bursty and synchronized, while inference requires fast and consistent responses and agentic systems add distributed activity across services, tools and data sources. AMD said those patterns can reduce flow diversity, and that a small number of congested paths can affect the performance of an entire training job and reduce utilization of expensive GPUs.

The Oracle post said the AMD portfolio on OCI spans EPYC processors and Instinct MI300X, MI355X and MI455X GPUs. Oracle launched MI300X-based OCI instances in 2024 and announced general availability of OCI Compute with AMD Instinct MI355X GPUs in October 2025.