Oracle deployed a mega-scale AI supercomputer, establishing itself as a major infrastructure builder alongside cloud hyperscalers.

Oracle has announced OCI Zettascale10, which the company is describing as the "largest AI supercomputer in the cloud." The system will deliver up to 16 zettaFLOPS of peak performance and forms the foundation of the Stargate supercluster that Oracle is building with OpenAI in Abilene, Texas. The Zettascale10 represents a significant expansion from Oracle's first Zettascale cloud computing cluster, announced at Cloud World 2024, which promised up to 131,072 Nvidia GPUs—the new system will scale up to 800,000 Nvidia GPUs, representing roughly 6x more capacity than its predecessor.

To achieve this scale while maintaining efficiency, OCI has combined Oracle Acceleron RoCE networking architecture with Nvidia's AI infrastructure to deliver "extremely low" GPU-to-GPU latency. The Zettascale10 clusters will be housed in large gigawatt data center campuses located within a 2km radius for reduced latency. According to OCI EVP Mahesh Thiagarajan, "Customers can build, train, and deploy their largest AI models into production using less power per unit of performance and achieving high reliability."

Oracle's broader commitment to this infrastructure expansion is substantial. The company has committed 4.5 gigawatts of additional Stargate capacity at a cost of over $300 billion, with new sites in Texas, New Mexico and the Midwest also in the works. Oracle's co-CEO Clay Magouyrk stated, "To meet this enormous demand, we continue to expand OCI's footprint at an unrivaled pace to deliver the most performant and cost-effective AI training and inferencing." The Zettascale10 is set to become available in the second half of 2026.