Oracle just dropped a bombshell at Oracle AI World, unveiling its new OCI Zettascale10 computing cluster powered by NVIDIA GPUs that delivers up to 16 zettaflops of peak AI performance. This isn't just another cloud announcement - it's a direct challenge to hyperscale competitors like Amazon and Microsoft, promising to interconnect millions of GPUs with unprecedented efficiency. The partnership expands far beyond raw compute, integrating NVIDIA's entire AI software stack directly into Oracle's database and cloud platform.
The enterprise AI battle just got a major new combatant. Oracle stunned the industry today at Oracle AI World by announcing its OCI Zettascale10 computing cluster, a massive NVIDIA GPU-powered system delivering up to 16 zettaflops of peak AI compute performance. That's enough raw power to train the largest language models and run inference at unprecedented scale.
But Oracle isn't just throwing silicon at the problem. The company is using NVIDIA's Spectrum-X Ethernet - the first networking platform purpose-built for AI - to interconnect millions of GPUs with efficiency that traditional InfiniBand can't match. "This is a game-changer for hyperscale AI," said Mahesh Thiagarajan, Oracle's executive vice president of Cloud Infrastructure, during the announcement.
The timing couldn't be more strategic. While Amazon Web Services and Microsoft Azure dominate cloud AI infrastructure, Oracle is betting that its deep enterprise relationships and database expertise give it a unique advantage. The company isn't just offering raw compute - it's integrating AI directly into the data layer where most enterprise AI applications will live.
Oracle Database 26ai, the company's flagship database, now supports NVIDIA NeMo Retriever integration, allowing developers to build retrieval-augmented generation pipelines using NVIDIA NIM microservices without leaving the database environment. "We're eliminating the complexity of moving data between systems for AI workloads," explained Ian Buck, NVIDIA's vice president of hyperscale computing.
The partnership runs deeper than hardware announcements suggest. Oracle's new AI Data Platform includes built-in NVIDIA GPU options and features the NVIDIA RAPIDS Accelerator for Apache Spark, enabling GPU acceleration for Spark applications without code changes. This matters because most enterprise data processing still happens in Spark environments.
Oracle Media and Entertainment is already using NVIDIA's NeMo Curator library with Nemotron vision language models to automate video processing workflows. The system handles video decoding, segmentation, and transcoding while generating dense captions - the kind of multimodal AI processing that's becoming essential for modern applications.
