AI is remodeling the way in which enterprises construct, deploy and scale clever functions. As demand surges for enterprise-grade AI functions that supply velocity, scalability and safety, industries are swiftly shifting towards platforms that may streamline knowledge processing and ship intelligence at each layer of the enterprise.
At Oracle AI World, Oracle at this time introduced a brand new OCI Zettascale10 computing cluster accelerated by NVIDIA GPUs, designed for high-performance AI inference and coaching workloads. The cluster will ship as much as 16 zettaflops of peak AI compute efficiency and harness NVIDIA Spectrum-X Ethernet — the primary Ethernet platform purpose-built for AI — enabling hyperscalers to interconnect tens of millions of GPUs with unprecedented effectivity and scale.
Different bulletins embody added assist for NVIDIA NIM microservices in Oracle Database 26ai, NVIDIA accelerated computing integration within the new Oracle AI Knowledge Platform, native availability of the NVIDIA AI Enterprise software program platform within the OCI Console and extra.
“I imagine the AI market has been outlined by vital partnerships such because the one between Oracle and NVIDIA,” stated Mahesh Thiagarajan, government vice chairman of Oracle Cloud Infrastructure. “These partnerships present drive multipliers that assist guarantee buyer success on this quickly evolving house. OCI Zettascale10 delivers multi‑gigawatt capability for probably the most difficult AI workloads with NVIDIA’s next-generation GPU platform. As well as, the native availability of NVIDIA AI Enterprise on OCI provides our joint clients a number one AI toolset shut at hand to OCI’s 200+ cloud companies, supporting an extended tail of buyer innovation.”
“Via this newest collaboration, Oracle and NVIDIA are marking new frontiers in cutting-edge accelerated computing — streamlining database AI pipelines, rushing knowledge processing, powering enterprise use instances and making inference simpler to deploy and scale on OCI,” stated Ian Buck, vice chairman of hyperscale and high-performance computing at NVIDIA.
Rushing AI Database Workloads
Oracle Database 26ai, Oracle’s flagship database, is including key performance to speed up high-volume AI vector workloads.
Oracle Database 26ai utility programming interfaces now assist integration with NVIDIA NeMo Retriever, permitting builders to simply run vector embedding fashions or implement retrieval-augmented technology (RAG) pipelines utilizing NVIDIA NIM microservices.
NVIDIA affords a full suite of NIM microservices for each stage of a RAG pipeline: NeMo Retriever extraction fashions for ingesting multimodal knowledge at scale, NeMo Retriever embedding fashions for changing knowledge chunks into vector embeddings, NeMo Retriever reranking fashions for enhancing total accuracy of ultimate responses, and enormous language fashions (LLMs) to generate the ultimate contextually correct responses.
Oracle Non-public AI Providers Container is a brand new service that makes it straightforward to deploy AI companies wherever wanted, together with cloud and on-premises environments. Oracle’s first implementation, which helps execution on CPU sources, has now been designed to assist the longer term use of NVIDIA GPUs for vector embedding and index technology utilizing the NVIDIA cuVS open-source library.
Embedding technology and vector search index creation are two essential duties required by vector databases. As knowledge volumes enhance and AI functions mature, vector index creation construct instances more and more grow to be a bottleneck.
GPUs are extremely environment friendly at constructing approximate nearest neighbor search algorithms. Customers who need to speed up their index builds will quickly be capable of offload this computationally intensive activity to NVIDIA GPUs utilizing deliberate Oracle Non-public AI Providers Container capabilities. As soon as the index has been constructed, it may be formatted for search on Oracle AI Database 26ai.
Oracle AI Knowledge Platform and NVIDIA RAPIDS Accelerator for Apache Spark
NVIDIA accelerated computing can be built-in into the brand new Oracle AI Knowledge Platform, which gives a complete ecosystem that unites enterprise knowledge with AI fashions, developer instruments, and tight controls over privateness and governance.
The Oracle AI Knowledge Platform features a built-in NVIDIA GPU choice to energy high-performance workloads. It additionally contains a new NVIDIA RAPIDS Accelerator for Apache Spark plug-in to unlock sooner analytics, extract, remodel, load, and machine studying pipelines by GPU acceleration.
The RAPIDS Accelerator for Apache Spark plug-in makes use of GPUs to speed up processing by combining the facility of the NVIDIA cuDF library and the size of the Spark distributed computing framework. All of that is designed to allow GPU-acceleration for Apache Spark functions with no code modifications.
Powering Enterprise AI Functions With NVIDIA Nemotron and NeMo
Oracle Media and Leisure is utilizing the NVIDIA NeMo Curator library with a Nemotron imaginative and prescient language mannequin (VLM) to energy video understanding.
This pipeline accelerates Oracle’s video-centric AI workflows by automating the pre-processing steps: video decoding, clip segmentation, transcoding and extra. It permits high-quality, scalable filtering, deduplication, annotation, classification and high quality management for each video and related textual content. This functionality permits Oracle to generate dense video captions and curate photographs wanted to coach downstream fashions, enhancing their effectivity and reliability.
NVIDIA NeMo Retriever Parse, a transformer-based vision-encoder-decoder mannequin designed for high-precision doc understanding, enhances Oracle Fusion Doc Intelligence by making it simpler to extract significant info from complicated paperwork. The mannequin goes past easy textual content scanning — it could deal with versatility, range and variability in enterprise paperwork, extracting vital metadata whereas preserving doc construction. These capabilities can be utilized to construct agentic or multimodal RAG functions.
Bringing all these capabilities collectively, Oracle AI Hub now affords enterprises a single entry level for constructing, deploying and managing customized AI options.
Customers can deploy NVIDIA NIM microservices by Oracle AI Hub, delivering a easy, no-code expertise for deploying fashions, together with NVIDIA Nemotron LLMs, VLMs and extra. The preliminary launch contains a curated set of hosted NIM microservices and early entry to next-generation, streamlined inference capabilities. With the mixing of NIM microservices, designed to run a broad vary of LLMs from a single container, clients can shortly deploy fashions for varied enterprise functions.
NVIDIA AI Enterprise on OCI
Enterprises can now additionally harness NVIDIA AI Enterprise, natively built-in inside OCI, for simplified entry to NVIDIA’s cloud-native suite of software program instruments, libraries and frameworks. This integration streamlines the event, deployment and administration of AI options, offering strong enterprise assist throughout Oracle’s platform.
NVIDIA AI Enterprise is now natively obtainable throughout the OCI Console expertise, permitting customers to instantly allow it when provisioning supported GPU cases. This functionality is out there throughout OCI’s distributed cloud, together with public areas, sovereign clouds and devoted areas, to assist clients meet safety and compliance necessities.
This new providing permits clients to entry a full suite of AI instruments with out having to individually procure orders by the Oracle Cloud Market, offering a streamlined course of to construct AI functions at scale with versatile pricing, enterprise assist, knowledgeable steerage and precedence safety updates.
NVIDIA was additionally acknowledged at Oracle AI World as a 2025 Oracle Companion award winner, underscoring the corporate’s work with Oracle to remodel the AI panorama for enterprises and drive innovation throughout the OCI ecosystem.
To study extra, learn the Oracle press launch.