Oracle Cloud Infrastructure has been named as an official NVIDIA Exemplar Cloud, validating OCI’s ability to deliver consistent, high-performance AI infrastructure at scale. Oracle is one of the first hyperscalers to earn this designation with NVIDIA Blackwell GPUs.

The Exemplar Cloud initiative provides customers with a transparent yardstick for comparing cloud performance on NVIDIA’s accelerated computing infrastructure with reproducible engineering results.

Why It Matters

To be validated as an Exemplar Cloud, each provider must demonstrate that their infrastructure can run NVIDIA’s reference workloads within 5% of NVIDIA’s target results, ensuring reproducible outcomes and predictable scaling.

On OCI, the most recent results come from the BM.GPU.B200.8 compute shape – the same infrastructure powering customer clusters with tens of thousands of GPUs.

Each bare-metal node includes:

  • 8 x NVIDIA Blackwell GPUs with 180GB of memory per GPU
  • 2 x 64 Intel 8592+ CPU cores
  • 4TB of CPU memory
  • 8 x 3.84 TB of local NVMe storage
  • 2 x 200 Gbps front-end NICs
  • 8 x 400 Gbps RDMA NICs

OCI’s supercluster architecture enables these GPUs to operate as a single system, maintaining predictable latency and throughput from rack one to rack one thousand.

Performance Customers Can Reproduce

NVIDIA’s benchmarks are public, and customers can reproduce them directly on OCI. Whether running NCCL communication tests, training multi-node models, or scaling inference workloads, customers will see the same performance that NVIDIA validated.

This transparency gives customers a trusted baseline for comparing performance across clouds.

Partnership in Practice

“Exemplar Cloud validation highlights the engineering collaboration between NVIDIA and Oracle,” said Alexis Black Bjorlin, Vice President/General Manager of DGX Cloud at NVIDIA. “OCI’s results on the NVIDIA Blackwell platform show what’s possible when our teams co-design for scale that is reproducible performance across tens of thousands of GPUs.”

Oracle’s pursuit of Exemplar Cloud validation builds on years of collaborative engineering with NVIDIA. Teams across both companies – from networking to architecture and solutions – have collaborated to ensure OCI’s AI infrastructure performs reliably under sustained, real-world loads.

This collaboration extends beyond NVIDIA HGX B200 to NVIDIA GB200 NVL72 and even the upcoming NVIDIA Blackwell Ultra platforms arriving later this year. OCI’s goal is to keep every generation of GPU infrastructure benchmarked, measurable, and production-ready the moment it launches.

Looking Ahead

“Exemplar Cloud validates what our customers already experience, OCI’s GPU infrastructure accelerated by NVIDIA delivers consistent, reproducible performance at scale. It gives every customer confidence that what we publish is what they’ll experience,” said Ben Scarborough, SVP of Customer Engineering for OCI.

The NVIDIA Exemplar Cloud milestone reinforces Oracle’s leadership in AI infrastructure. OCI will continue scaling superclusters globally and applying the same engineering rigor to each new generation of hardware.

Closing Thought

At OCI we build infrastructure that enables customers to push the limits of what’s possible.

Achieving NVIDIA Exemplar Cloud with NVIDIA HGX B200 proves that when Oracle and NVIDIA partner, we raise the bar for cloud-scale AI.