Oracle CloudWorld conference attendees witness Oracle Cloud Infrastructure unveil its first zettascale OCI Supercluster.
At the Oracle CloudWorld conference, Oracle Cloud Infrastructure (OCI) unveiled its first zettascale OCI Supercluster, powered by the NVIDIA Blackwell platform. This solution is created to help enterprises accelerate AI workloads and data processing, leveraging over 100,000 of NVIDIA’s latest-generation GPUs. These Superclusters offer flexibility for deployment, whether on-premises, in public, or sovereign cloud environments.
In its maximum scale, the OCI Supercluster offers more than three times as many GPUs as the Frontier supercomputer and more than six times that of other hyperscalers. These systems can scale to 131,072 Blackwell GPUs and deliver 2.4 zettaflops of peak AI compute, enabled by NVIDIA ConnectX-7 NICs and Quantum-2 InfiniBand networking. The Blackwell-based systems are expected to be available in the first half of 2025.
NVIDIA GB200 NVL72 liquid-cooled bare-metal instances previewed
Oracle also previewed its NVIDIA GB200 NVL72 liquid-cooled bare-metal instances at the conference, which are designed for generative AI applications. These instances enable large-scale AI training and real-time inference of trillion-parameter models, acting as a single, massive GPU through the expanded 72-GPU NVIDIA NVLink domain. Additionally, OCI announced its upcoming availability of NVIDIA HGX H200 Tensor Core GPUs, capable of connecting eight GPUs in a single bare-metal instance, allowing enterprises to scale up to 65,536 H200 GPUs. These offerings are expected to accelerate real-time AI inference and training at scale.
Introducing NVIDIA L40S GPU-accelerated instances
For midrange AI workloads, Oracle introduced the general availability of NVIDIA L40S GPU-accelerated instances. Oracle’s edge solutions also provide scalable AI deployments, including the Roving Edge Device v2, which supports up to three NVIDIA L4 Tensor Core GPUs, even in remote or disconnected locations.
Enterprises are already leveraging OCI Superclusters for AI innovation. Reka, a foundation model startup, is using these clusters to develop advanced multimodal AI models for enterprise agents. According to Dani Yogatama, cofounder and CEO of Reka, the infrastructure allows them to handle large models and contexts efficiently while scaling training at the cluster level.
Oracle and NVIDIA also showcased new integrations to accelerate generative AI workloads on Oracle Autonomous Database. These integrations highlight how NVIDIA GPUs can enhance AI capabilities in Oracle databases. The demonstrations included accelerating bulk vector embeddings, optimizing vector graph index generation, and boosting text generation and translation with NVIDIA NIM inference microservices. These innovations allow enterprises to seamlessly integrate AI into their structured and unstructured data management within Oracle databases.
Addressing data residency
On a global scale, Oracle and NVIDIA are collaborating to deliver sovereign AI infrastructure, addressing data residency needs for governments and enterprises. Brazil-based startup Wide Labs used NVIDIA H100 Tensor Core GPUs and the NeMo framework in OCI’s Brazilian data centers to create Amazônia IA, a large language model for Brazilian Portuguese, ensuring data sovereignty. Similarly, Nomura Research Institute in Japan is enhancing its financial AI platform with LLMs while adhering to financial regulations and data sovereignty requirements using OCI’s Alloy infrastructure with NVIDIA GPUs. Zoom is also leveraging NVIDIA GPUs in OCI’s Saudi Arabian data centers to comply with local data regulations.
Geospatial modeling company RSS-Hydro is using NVIDIA-powered OCI infrastructure to simulate flood impacts in Japan’s Kumamoto region. This demonstrates how AI can be applied to climate change mitigation efforts through digital twin technology.
Enterprises can also accelerate task automation on OCI using NVIDIA software such as NIM microservices and NVIDIA cuOpt. These solutions enable businesses to quickly adopt generative AI for tasks like code generation and route optimization. NVIDIA’s AI Enterprise software platform, which includes these tools, is available on the Oracle Cloud Marketplace, providing enterprises with powerful resources to enhance AI-driven workflows and agentic processes.
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed
