Oracle and AMD have today announced a major extension of their long‑standing, multi‑generation partnership aimed at helping customers scale their AI capabilities significantly. From calendar Q3 2026, Oracle will become the first hyperscaler to offer a publicly accessible AI supercluster powered by 50,000 AMD Instinct MI450 Series GPUs, with further expansion planned for 2027 and beyond.
The move builds on several years of joint development, beginning with the launch of AMD Instinct MI300X shapes on Oracle Cloud Infrastructure (OCI) in 2024 and the subsequent rollout of OCI Compute with AMD Instinct MI355X GPUs. These offerings feed into the broader zettascale OCI Supercluster.
The new AI superclusters will employ AMD’s “Helios” rack design, integrating AMD Instinct MI450 GPUs, the upcoming “Venice” EPYC CPUs, and Pensando advanced networking codenamed “Vulcano.” This vertically optimised rack‑scale architecture is designed to deliver high performance, energy efficiency and scalability for large‑scale AI training and inference workloads.
Mahesh Thiagarajan, Executive Vice President, Oracle Cloud Infrastructure, said: “Our customers are building some of the world’s most ambitious AI applications, and that requires robust, scalable, and high‑performance infrastructure. By bringing together the latest AMD processor innovations with OCI’s secure, flexible platform and advanced networking powered by Oracle Acceleron, customers can push the boundaries with confidence. Through our decade‑long collaboration with AMD—from EPYC to AMD Instinct accelerators—we’re continuing to deliver the best price‑performance, open, secure, and scalable cloud foundation in partnership with AMD to meet customer needs for this next era of AI.”
Forrest Norrod, Executive Vice President and General Manager of AMD’s Data Center Solutions Business Group, added: “AMD and Oracle continue to set the pace for AI innovation in the cloud. With our AMD Instinct GPUs, EPYC CPUs, and advanced AMD Pensando networking, Oracle customers gain powerful new capabilities for training, fine‑tuning, and deploying the next generation of AI. Together, AMD and Oracle are accelerating AI with open, optimised, and secure systems built for massive AI data centres.”
The MI450 GPU‑powered shapes are engineered to support sophisticated workloads including generative AI and high‑performance computing. Each GPU delivers up to 432 GB of HBM4 memory and 20 TB/s of memory bandwidth, enabling users to train and infer models 50 percent larger entirely in‑memory. The “Helios” rack design allows dense, liquid‑cooled, 72‑GPU racks and combines UALoE scale‑up connectivity with Ethernet‑based scale‑out networking for low latency and high throughput across GPU clusters.
In addition, the supercluster architecture features advanced head nodes based on EPYC “Venice” CPUs offering confidential computing and security features. AMD Pensando DPUs enable converged networking that provides line‑rate data ingestion, improves performance, and enhances security. The design allows GPUs to be equipped with up to three 800 Gbps AI‑NICs (Vulcano), supporting lossless, high‑speed programmable connectivity with standards such as RoCE and UEC.
The architecture also incorporates the open UALink and UALoE fabric, which simplifies data sharing across GPUs within racks without routing through CPUs. The open‑source ROCm software stack facilitates vendor flexibility and supports popular AI and HPC frameworks, while advanced partitioning and virtualization features enable secure multi‑tenant resource sharing and fine‑grained allocation of GPU workloads.
Alongside the MI450 deployment, Oracle also announced the general availability of OCI Compute instances powered by AMD Instinct MI355X GPUs. These will be part of the zettascale OCI Supercluster capable of scaling to 131,072 GPUs, offering customers alternative optimisation and flexibility for AI workloads.
