Oracle and AMD Announce Massive AI Expansion
Oracle and AMD have significantly expanded their partnership. The goal is to help customers scale their Artificial Intelligence capabilities. Oracle Cloud Infrastructure (OCI) will be a launch partner. It will host the first publicly available AI supercluster. This powerful cluster will be powered by AMD Instinct MI450 Series GPUs. The initial deployment is substantial. It involves 50000 GPUs beginning in the third quarter of 2026. Further expansion is already planned for 2027 and beyond. This collaboration is a major milestone for AMD. It strengthens its position in the competitive AI hardware market significantly. It also offers a strong alternative to the dominant chips from rivals.
Next Generation Helios Rack Design Features
The new OCI AI superclusters will use the AMD “Helios” rack design. This new architecture is vertically optimized. It is built specifically for massive AI data centers. It incorporates several advanced components. The system features the new AMD Instinct MI450 Series GPUs. It also includes the next generation AMD EPYC CPUs. These CPUs are codenamed “Venice”. Finally, it utilizes the next generation AMD Pensando advanced networking. The networking is codenamed “Vulcano”. This system is designed to maximize cluster use. It also streamlines large scale AI workflows considerably.
Instinct MI450 GPU Delivers Exceptional Performance
The upcoming AMD Instinct MI450 Series accelerators offer cutting edge features. Each GPU will provide up to 432 GB of HBM4 memory. It also boasts 20 TB/s of memory bandwidth. This massive capacity allows for significant model handling. Customers can train and perform inference on models. These models are 50 percent larger than previous generations. They can also be run entirely in memory. Furthermore, the system uses AMD’s standards. These are the UALink and UALoE interconnects. These standards ensure high speed communication between GPUs. This hardware coherence reduces latency. It also minimizes CPU routing overhead. Each MI450 GPU can also link up. It connects with up to three 800 Gbps Pensando “Vulcano” AI NICs. This provides lossless, high speed, and programmable connectivity. The “Venice” EPYC CPUs will also offer confidential computing capabilities. This further enhances security and data processing speed.
Strategic Deployment for the Cloud Market
This large-scale deployment builds on Oracle’s existing AMD offerings. OCI currently uses Instinct MI300X and MI355X GPU instances. The decision to deploy 50000 MI450 GPUs is strategic. It supports Oracle’s goal to compete fiercely. It aims to rival major cloud providers like Microsoft, Amazon, and Google. The new systems will allow OCI to target massive AI compute power. They are projected to deliver 1 point 4 exaFLOPS of performance. This new offering will be available in the second half of 2026. Oracle is already taking orders for its OCI Zettascale10 service. This new capacity is crucial for the future of AI. It gives Southeast Asia customers a powerful new option for their AI development.
