Oracle and AMD have decided to take their longstanding collaboration to a new level this year. This expanded partnership will enable Oracle to deploy 50,000 AMD Instinct MI450 Series GPUs in a new data center. The rollout will begin in the third calendar quarter of 2026 and expand throughout 2027 and beyond. Oracle will thus become the first hyperscaler to offer a publicly available AI supercluster powered by these AMD GPUs. The collaboration builds on years of joint innovation, with Oracle Cloud Infrastructure (OCI) already using AMD Instinct GPU platforms since 2024, including the MI300X and now the MI355X.
This initiative responds to growing demand for large-scale AI infrastructure. Customers need a flexible and open solution for training and running advanced AI models that exceed the limits of current clusters. The new OCI AI superclusters will be powered by AMD's rack design called “Helios,” which includes AMD Instinct MI450 Series GPUs, AMD EPYC processors code-named “Venice,” and advanced AMD Pensando networking code-named “Vulcano.” This vertically optimized system delivers high performance, scalability, and energy efficiency for large-scale AI workloads.
Benefits of AMD GPUs
AMD Instinct MI450 Series GPUs bring a range of advanced features to OCI. Each GPU offers up to 432 GB of HBM4 memory and 20 TB/s of memory bandwidth, allowing models 50% larger than those supported by the previous generation to be processed directly in memory. This accelerates results and handles more complex workloads without the need to partition models. The “Helios” rack design supports dense, liquid-cooled racks with 72 GPUs, optimizing performance, cost, and power consumption.
The cluster head is equipped with AMD EPYC “Venice” processors, which increase cluster utilization and accelerate workload orchestration. These processors also provide confidential computing and built-in security features to protect sensitive AI data. DPU-accelerated (Data Processing Unit) networking based on AMD Pensando ensures high-speed data ingestion, improves performance, and strengthens security for large AI and cloud infrastructures.
Each GPU can be equipped with up to three 800 Gbps AMD Pensando “Vulcano” AI-NICs, providing lossless, high-speed connectivity that supports RoCE and Ultra Ethernet Consortium (UEC) standards. This enables fast distributed training and optimized collective communications.
AMD ROCm software, which is open-source, enables rapid innovation and easy migration of AI and HPC (high-performance computing) workloads. It supports popular frameworks, libraries, compilers, and runtime environments. Advanced GPU partitioning and virtualization enable secure cluster sharing, needs-based GPU allocation, and robust multi-tenancy (sharing among multiple users).
OCI also announced the general availability of OCI Compute with AMD Instinct MI355X GPUs, which can scale up to 131,072 GPUs in a zettascale OCI Supercluster. These GPUs offer excellent value, cloud flexibility, and open-source compatibility.
Executive Perspectives
“Our customers are building some of the world’s most ambitious AI applications, and that requires robust, scalable, and high-performance infrastructure,” said Mahesh Thiagarajan, executive vice president, Oracle Cloud Infrastructure. “By combining AMD’s latest processor innovations with OCI’s secure and flexible platform and advanced Oracle Acceleron networking, customers can confidently push the boundaries.”
“AMD and Oracle continue to set the pace for AI innovation in the cloud,” added Forrest Norrod, executive vice president and general manager of the Data Center Solutions Business Group at AMD. “With our AMD Instinct GPUs, EPYC processors, and advanced AMD Pensando networking, Oracle customers gain powerful new capabilities for training, fine-tuning, and deploying the next generation of AI.”
This agreement underscores AMD’s growing role as an alternative to Nvidia in the AI chip market. Oracle has been working with AMD for several years and plans to offer the next generation of AMD GPUs in the second half of next year. Competition in the AI chip sector includes companies such as Broadcom and Cerebras, but this partnership strengthens AMD’s position in massive AI data centers.
Sources: oracle.com and theinformation.com
