The latest generation of Nvidia hardware is entering general availability on Oracle Cloud Infrastructure.

Announced during the Oracle AI World event, the Nvidia Vera Rubin NVL72 systems and Vera CPUs are now generally available on Oracle’s cloud platform.

Oracle is also set to add the Nvidia Vera Rubin systems to its “Superclusters” offering, which will be a multi-gigawatt-scale deployment. At a later date, the Rubin HGX NVL8 Systems will also be added, which will offer eight Rubin GPUs connected through NVLink in an x86-based server configuration.

OCI Superclusters combine Nvidia NVL72 rack-scale systems, Oracle Acceleron networking with Nvidia Spectrum-X Ethernet networking featuring Nvidia Spectrum-6 switches and Nvidia BlueField-4 DPUs, and gigawatt-class power and cooling to train and serve the largest AI models. Each rack connects 72 GPUs through Nvidia NVLink scale-up networking, while OCI’s resilient, multiplanar network scales across hundreds of thousands of GPUs.

The addition of Vera Rubin to Superclusters builds on Oracle’s existing Nvidia Grace Blackwell compute that is available in the formation and was used in Abilene by OpenAI to train GPT-6 Astra.

OpenAI used an OCI Supercluster of more than 100,000 Blackwell GPUs.

“As our frontier models advance, we need infrastructure that can keep pace,” said Sreekanth Pothanis, engineering leader, compute infra at OpenAI. “Nvidia Vera Rubin on OCI gives us the computing performance and scale to accelerate model training and serve increasingly demanding AI workloads.”

The deployment of Vera CPUs alone is targeting agentic AI workloads, combining high single-thread performance with memory bandwidth to help agents complete CPU-intensive tasks faster.

“Oracle Cloud Infrastructure gives customers the performance, security and scale to advance frontier AI,” said Mahesh Thiagarajan, president of Oracle Cloud Infrastructure, Oracle. “Our collaboration with Nvidia extends those capabilities across training, inference and agentic AI. By combining Nvidia Vera Rubin GPUs and Vera CPUs with Oracle Acceleron networking and AI-optimized data centers, we’re helping customers keep GPUs productive, secure, and scale efficiently as their needs grow.”

Oracle goes CPU crazy

Beyond Nvidia hardware, Oracle has announced a multi-year collaboration with Intel to expand Intel Xeon-based computing on OCI, including current Xeon platforms and the planned OCI X14 offerings based on Intel’s next-generation server processor codenamed Diamond Rapids.

CPUs are becoming increasingly important to AI workloads as AI agents are being rapidly adopted. The partnership with Intel will increase CPU capacity on OCI, while giving customers access to familiar tooling and software environments.

“As AI agents scale, they need processors that can run software, move and secure data, and accelerate inference. Xeon brings together performance and hardware-based confidential computing, giving organizations a trusted, cost-effective foundation for AI at scale,” said Kevork Kechichian, executive vice president and general manager, Intel Data Center Group. “Our expanded collaboration with Oracle is intended to combine Intel's multigenerational roadmap with Oracle Cloud Infrastructure’s technology to help customers put AI to work faster and more cost-effectively.”

Planned Diamond Rapids-based OCI X14 offerings will target both AI and general-purpose computing, including Intel Advanced Matrix Extensions (AMX) with FP8, FP16, and BF16 support for inference, and Intel Advanced Performance Extensions (APX) to help applications execute with less instruction overhead.

Meanwhile, with AMD, Oracle is nearing bringing online its E7 instances that will use sixth-generation AMD Epyc processors, with AMD Pensando DPUs supporting Oracle Acceleron NICs for AMD-powered instances.

The E7 instances are designed to support CPU work surrounding AI models. The instances use 6th Gen AMD Epyc processors from the Epyc 9006 Server Processor Series. 6th Gen Epyc processors provide up to 256 cores and 512 threads per socket, up to 1,024 MB of L3 cache, 16 memory channels with up to 8000 MTps, and PCIe 6.0 connectivity. These provide significant uplifts over 5th Gen Epyc CPUs, and the company says they power enterprise workloads more quickly and efficiently, and allow for the deployment of more agents per instance and per watt.

Oracle has further revealed plans to bring a variety of chip architectures onto its platform, including Cerebras’ wafer-scale compute, SambaNova’s chips, and more.

The Cerebras chips are designed for high-speed AI inference, and integrate compute and memory on a single wafer-scale processor to reduce data movement and inter-chip communication.

“People want AI that can do valuable work quickly,” said Andrew Feldman, co-founder and CEO of Cerebras. “Every step an agent makes requires inference, whether it’s reasoning, checking its work, or taking action. The faster the inference, the more work AI can do without making users wait. Bringing Cerebras fast inference to Oracle Cloud Infrastructure gives AI labs, enterprises, and developers the speed to build agents that do more, faster.”

Cerebras Inference on OCI is planned to be available in calendar year 2027.

Details about the SambaNova partnership remain sparse, beyond that Oracle will bring online the SN50 chip from the company. No specific timeline has yet been provided.

The final player is the Positron Atlas chip, which is now generally available on OCI.

Positron Atlas Inference Servers are purpose-built for Large Language Models, and deliver low-latency performance for popular open-weight models. Positron Atlas is networked through Oracle Acceleron.

Atlas is Positron's first-generation chip, was fabricated by Intel in the US, and is currently shipping to customers. Positron claims the hardware can achieve three times the compute-per-watt of Nvidia’s H100 GPUs and support up to 500-billion-parameter models.

The new chips coming to OCI join the Google TPU, which Oracle today revealed was set to be deployed on OCI under a multi-year and multi-gigawatt contract.