Oracle launches first zettascale supercluster with NVIDIA tech

Oracle Cloud Infrastructure (OCI) has announced the launch of its first zettascale OCI Supercluster, accelerated by NVIDIA's Blackwell platform, aimed at supporting enterprises in training and deploying next-generation AI models. The announcement, made at the Oracle CloudWorld conference, highlights the integration of over 100,000 of NVIDIA's latest-generation GPUs to help businesses enhance their AI capabilities.

The new OCI Superclusters offer flexibility for customers by providing a range of NVIDIA GPUs that can be deployed on-premises, in public cloud settings, or in sovereign cloud environments. These Blackwell-based systems are expected to be available in the first half of next year. They can scale up to 131,072 Blackwell GPUs, using NVIDIA ConnectX-7 NICs for RoCEv2 or NVIDIA Quantum-2 InfiniBand networking. The setup aims to deliver up to 2.4 zettaflops of peak AI compute to the cloud.

In addition to the superclusters, Oracle previewed NVIDIA GB200 NVL72 liquid-cooled bare-metal instances to enhance generative AI applications. These instances will possess the capability for large-scale training utilising Quantum-2 InfiniBand and offer real-time inference of trillion-parameter models within the expanded 72-GPU NVIDIA NVLink domain, which can function as a single extensive GPU.

This year, OCI will also provide NVIDIA HGX H200 – connecting eight NVIDIA H200 Tensor Core GPUs in a single bare-metal instance via NVLink and NVLink Switch, and scaling to 65,536 H200 GPUs with NVIDIA ConnectX-7 NICs over RoCEv2 cluster networking. These instances are aimed at customers seeking to deliver real-time inference at scale and expedite their training workloads.

OCI has also unveiled the general availability of NVIDIA L40S GPU-accelerated instances for midrange AI workloads, NVIDIA Omniverse, and visualisation. Oracle's edge offerings now support scalable AI at the edge with acceleration by NVIDIA GPUs, even in disparate and remote locations, including deployments with Oracle's Roving Edge Device v2 supporting up to three NVIDIA L4 Tensor Core GPUs.

Enterprises are leveraging NVIDIA-powered OCI Superclusters to drive AI innovation. For instance, Reka, a foundation model startup, is using the clusters to develop advanced multimodal AI models aimed at creating enterprise agents. Dani Yogatama, cofounder and CEO of Reka, stated, "Reka's multimodal AI models, built with OCI and NVIDIA technology, empower next-generation enterprise agents that can read, see, hear, and speak to make sense of our complex world. With NVIDIA GPU-accelerated infrastructure, we can handle very large models and extensive contexts with ease, all while enabling dense and sparse training to scale efficiently at cluster levels."

NVIDIA has been recognised with the 2024 Oracle Technology Solution Partner Award in Innovation for its comprehensive approach to innovation. Oracle Autonomous Database is also set to gain NVIDIA GPU support for Oracle Machine Learning notebooks, facilitating the acceleration of data processing workloads on Oracle Autonomous Database.

At the event, NVIDIA and Oracle are demonstrating capabilities showing how the NVIDIA accelerated computing platform can enhance key components of generative AI retrieval-augmented generation pipelines. The demonstrations involve accelerating bulk vector embeddings from Oracle Autonomous Database Serverless, using NVIDIA GPUs and cuVS to expedite vector graph index generation, and showcasing NVIDIA NIM microservices to boost performance in generative AI tasks such as text generation and translation.

NVIDIA and Oracle's collaboration extends to delivering sovereign AI infrastructure globally for addressing data residency needs. For example, Brazil-based Wide Labs utilised NVIDIA H100 Tensor Core GPUs and the NVIDIA NeMo framework in OCI's Brazilian data centres to develop Amazonia IA, a large language model for Brazilian Portuguese. "Developing a sovereign LLM allows us to offer clients a service that processes their data within Brazilian borders, giving Amazônia a unique market position," stated Nelson Leoni, CEO of Wide Labs.

Additionally, Japan's Nomura Research Institute is using OCI's Alloy infrastructure with NVIDIA GPUs to enhance its financial AI platform while maintaining compliance with financial regulations. Communication company Zoom will use NVIDIA GPUs in OCI's Saudi Arabian data centres to support local data requirements. Geospatial modeling company RSS-Hydro is demonstrating its flood mapping platform, powered by NVIDIA Omniverse and L40S GPUs in OCI, to mitigate climate change impacts in Japan's Kumamoto region.

Among numerous initiatives, nations and organisations worldwide are building and deploying domestic AI applications powered by NVIDIA and OCI to drive economic resilience through sovereign AI infrastructure.