Oracle and NVIDIA today announced expanded AI capabilities on Oracle Cloud Infrastructure (OCI) that help redefine scalable AI performance, accelerate vector database operations, and simplify enterprise AI deployment using cloud-native services. Together, Oracle and NVIDIA are enabling organizations to move from AI experimentation to production at extraordinary scale, speed, and efficiency.
Organizations training and serving frontier AI models require infrastructure engineered for extreme throughput, consistently ultra-low latency, and massive GPU scale. OCI Superclusters are built to meet that demand—engineered to connect hundreds of thousands of GPUs into a single AI supercomputer and support an unprecedented level of performance, with over 17 zettaFLOPS of peak performance, up to 131 Pb/s of cluster front-end throughput for massive scale-out, and up to 2.1 Eb/s of RDMA throughput with sub-10 microsecond latency for fast, efficient scale-in.
Next-generation AI Performance: OCI delivers nearly limitless supercomputing scale
At the core is the Oracle Acceleron network architecture, designed for predictable, high-bandwidth communication across large clusters. By combining RDMA over Converged Ethernet (RoCE), Converged Network Interface Card (CNIC) offload, and a multiplanar network design, Oracle Acceleron delivers deterministic performance and ultra-low-latency GPU-to-GPU connectivity at scale. This helps customers train larger models faster, run high-throughput inference more efficiently, accelerate multimodal and scientific workloads, and improve cluster utilization across Oracle’s distributed cloud.
Introducing the next OCI Supercluster—powered by NVIDIA Vera Rubin
Today, Oracle is introducing a next-generation OCI Supercluster powered by the NVIDIA Vera Rubin platform, including NVIDIA Rubin GPUs, NVIDIA Vera CPUs, NVIDIA BlueField-4 DPUs, sixth-generation NVLink, NVIDIA ConnectX-9 SuperNICs, and NVIDIA Spectrum-X Ethernet switches, purpose-built to accelerate next-generation training and high-throughput inference workloads.
The system integrates thousands of Rubin GPUs to deliver breakthrough AI compute performance at scale, while NVIDIA BlueField-4 DPUs and ConnectX-9 SuperNICs offload networking, security, and data movement from host CPUs to help increase throughput, improve workload isolation, and maximize usable GPU capacity across large-scale clusters.
Together, these technologies extend Oracle Acceleron’s multiplanar network architecture, using dedicated RoCE fabrics and direct GPU-to-GPU communication paths to reduce latency and increase bandwidth across thousands of nodes. By combining Oracle Acceleron’s deterministic networking with BlueField-4 data processing and Rubin GPU performance, OCI Superclusters delivers ultra-low-latency communication, higher cluster utilization, improved resilience across network planes, and optimized power efficiency at hyperscale.
From frontier model training to high-throughput inference and extreme-scale supercomputing, OCI Supercluster is built to help customers move faster, scale bigger, and push AI further.
Accelerating Oracle AI Database Embedding and Vector Index Creation
As organizations deploy retrieval-augmented generation (RAG) and AI-driven search applications, rapidly generating embeddings and maintaining large-scale vector indexes has become critical to delivering more accurate and responsive AI systems. Oracle AI Database can now use NVIDIA AI infrastructure and NVIDIA cuVS to accelerate large-scale embedding generation and vector index creation, helping reduce time-to-value for AI-driven applications.
With Oracle AI Database, developers can run vector similarity search directly in SQL or access capabilities through APIs and SDKs, allowing AI functionality to be integrated seamlessly into existing enterprise applications and workflows. GPU acceleration helps improve performance for embedding pipelines and indexing operations, enabling organizations to continuously refresh vector indexes as enterprise data grows.
Customers are leveraging Oracle AI Database with NVIDIA AI infrastructure and NVIDIA cuVS acceleration to power advanced knowledge retrieval, AI copilots, document intelligence, and real-time data exploration—unlocking faster insights and more efficient AI-driven operations.
OCI Generative AI Expands Open Model Innovation with NVIDIA Nemotron
OCI Generative AI continues to broaden how developers and enterprises use open-weights foundation models. Oracle Cloud Infrastructure recently announced support for the NVIDIA Nemotron 3 Super model through a new Model Import capability in OCI Generative AI.
Soon planned to be available on Oracle Government Cloud in addition to commercial cloud regions, NVIDIA Nemotron 3 Super is the first model from NVIDIA available through OCI Generative AI Model Import and demonstrates how organizations can continue to run advanced reasoning models on OCI while maintaining control over customization and deployment.
Model Import allows customers to bring supported models into OCI Generative AI and run them through the same managed service used for Oracle-hosted models. This combines the flexibility of open models with a consistent API, enterprise security model, and operational experience.
Oracle Government Cloud operates government cloud regions in the US, UK, and Australia and provides governments worldwide with a way to run generative AI models that still address local data residency, classification, operational, and security requirements.
These capabilities extend beyond infrastructure into enterprise applications. NVIDIA Nemotron models are now available to support Oracle Fusion Applications, augmenting generative AI-powered capabilities across business workflows such as finance, HR, supply chain, and customer experience. By combining Nemotron reasoning capabilities with Oracle’s enterprise application data, organizations can power intelligent automation, document understanding, and contextual decision-making directly within operational systems.
Oracle AI Database can access Nemotron models by calling NVIDIA NIM containers, enabling developers to build retrieval- augmented generation (RAG) applications using built-in vector search, embeddings, and AI-powered data processing. Together, OCI Generative AI, Oracle AI Database, and NVIDIA Nemotron models create a unified approach where enterprises can build AI applications that securely combine foundation models with enterprise data—helping accelerate development of intelligent applications across industries including financial services, healthcare, telecommunications, and media.
Conclusion
With these advancements, Oracle and NVIDIA continue to push the boundaries of enterprise AI—delivering supercomputing-scale performance, accelerated vector intelligence, and simplified cloud-native AI innovation in one integrated platform.
Learn more about additional NVIDIA + Oracle activities: