Oracle and AMD Partner to Boost AI Performance with New GPU Supercomputer

Austin, Texas and Santa Clara, Calif.—Jun 12, 2025 — Oracle and AMD announced a collaboration to integrate AMD Instinct™ MI355X GPUs into Oracle Cloud Infrastructure (OCI). This partnership aims to deliver breakthrough performance for large-scale AI and agentic workloads, with Oracle becoming one of the first hyperscalers to offer an AI supercomputer powered by these GPUs.

Key Highlights

• Zettascale AI Clusters: OCI will deploy clusters with up to 131,072 MI355X GPUs, enabling customers to build, train, and run AI models at unprecedented scale.
• Performance Gains: The MI355X GPUs offer 2.8X higher throughput and 50% more high-bandwidth memory than the previous generation, along with support for 4-bit floating point compute (FP4) for efficient inference.
• Liquid-Cooled Design: The dense, liquid-cooled racks maximize performance at 125 kilowatts per rack, with 64 GPUs per rack.
• Open-Source Compatibility: AMD ROCm software stack ensures flexibility and avoids vendor lock-in.
• Network Innovation: Oracle will be the first to deploy AMD Pollara™ AI NICs for advanced RoCE functionality and low-latency networking.

Quotes

• Mahesh Thiagarajan, EVP of Oracle Cloud Infrastructure: "AMD Instinct GPUs, paired with OCI’s performance, advanced networking, and scale, will help customers meet their AI workload demands."
• Forrest Norrod, EVP of AMD: "This collaboration provides customers with open solutions for high performance and efficiency as AI adoption grows."

Additional Resources

• Learn more about OCI AI infrastructure
• Explore OCI Compute
• Discover AMD GPUs on OCI

About Oracle

Oracle provides integrated cloud applications and autonomous infrastructure. Learn more at oracle.com.

Trademarks

AMD, AMD Instinct, Pollara, and ROCm are trademarks of Advanced Micro Devices, Inc.