Oracle Cloud Infrastructure (OCI) has taken a big step ahead in the global AI race by deploying thousands of NVIDIA Blackwell GPUs, representing a significant investment in the future of intelligent systems. This infrastructure, built specifically for agentic AI and advanced reasoning workloads, is intended to allow the construction of the most sophisticated AI models ever created—capable of planning, goal-setting, and autonomous decision-making.
Leading the Charge for Agentic AI and Reasoning Models
Oracle is pushing OCI to be the cloud of choice for agentic AI frameworks as AI progresses beyond text generation to systems capable of autonomous operation. These models underpin real-world applications like as autonomous robotics, next-generation virtual assistants, dynamic financial systems, and precision healthcare diagnostics.
To accommodate this increased demand, Oracle has begun to deploy NVIDIA GB200 Grace Blackwell Superchips throughout its data centres. These systems are designed for complicated, reasoning-based AI activities that demand massive compute power, memory capacity, and ultra-low latency networking.
What Makes the Blackwell GPU a Game changer?
NVIDIA’s Blackwell architecture marks a substantial jump in AI processing capabilities, providing:
- Massive performance increases in training and inference
- Improved energy efficiency enables sustainable AI at scale
- Built-in support for agentic AI, which involves real-time adaptability and long-term thinking
These features are important for enabling a new generation of models that not only understand language but also plan, act, and learn dynamically.
First-Phase Deployment: GB200 NVL72 Racks Go Live
Oracle has completed the first phase of the deployment of liquid-cooled NVIDIA GB200 NVL72 rack-scale systems, which are currently accessible to clients through NVIDIA DGX Cloud and Oracle Cloud Infrastructure. Each NVL72 unit consists of 36 Grace CPUs and 72 Blackwell GPUs, which are tightly integrated and engineered for maximum computation density and energy efficiency.
The racks use high-speed NVIDIA Quantum-2 InfiniBand and Spectrum-X Ethernet networking to provide scalable, low-latency connectivity suitable for huge AI model training and inference. This system is supported by Oracle and NVIDIA’s jointly designed software and database stack, which ensures optimal performance for real-world workloads.
Oracle’s vision: AI factories at cloud scale
This large-scale deployment represents a trend towards creating “AI factories” – next-generation data centres designed to generate intelligence at scale. These facilities allow organisations to process massive amounts of data using AI algorithms, transforming raw data into useful insights, autonomous behaviour, or real-time forecasts.
Oracle intends to expand its OCI Superclusters above 100,000 Blackwell GPUs, aligning with the global acceleration of AI innovation. Companies such as OpenAI and NVIDIA are already using Oracle infrastructure to train new reasoning models and create AI tools that power advanced applications across multiple industries.
A complete stack for scalable AI development
Oracle’s infrastructure is more than simply hardware; it provides a complete AI platform:
- Integrated support for NVIDIA AI Enterprise, CUDA, and common machine learning frameworks
- Native integration with the Oracle Autonomous Database, OCI Data Science, and Lakehouse services
- Deployment via NVIDIA DGX Cloud, including software, technical services, and full lifecycle support
Customers may now easily create, deploy, and manage high-performance AI workloads, whether in the public cloud or within a customer-controlled OCI Dedicated Region.
General Availability and Scalable Deployment Options
These capabilities are now widely available across all main OCI deployment options:
- Public Cloud
- Dedicated Region
- Oracle Alloy for partners’ and sovereign cloud use cases
Customers can grow from a single GPU instance to clusters of tens of thousands of GPUs based on workload needs. Oracle has already completed the initial phase of liquid-cooled NVIDIA GB200 NVL72 rack deployments, which are currently available to clients worldwide.
Why Oracle’s Move Matters
Oracle is one of the world’s first cloud providers heavily and early to deploy the GB200 NVL72 platform live after investing in Blackwell systems. The business intends to expand its OCI Superclusters beyond 100,000 GPUs, cementing its position as a serious player in AI cloud services alongside AWS, Azure, and Google Cloud.
Benefits for Oracle Cloud users include:
- Faster training and inference for the most complicated AI models
- Energy-efficient, rack-scale performance
- Broad industry application, from tech firms to national governments
Broad Industry Momentum and Real-World Applications
Interest in Oracle’s new AI infrastructure is high across sectors:
- Technology businesses are building LLMs and autonomous agents
- Government agencies are researching sovereign AI solutions
- Enterprise clients and regional cloud providers demand high-performance, low-latency solutions
Even NVIDIA intends to leverage Oracle-hosted systems to train models, build autonomous vehicles, and advance chip design.
Built for every AI scenario to unlock AI for Scale
Oracle’s infrastructure is capable of training the next generation of generative AI models, deploying enterprise copilots, and powering real-time AI applications. Oracle’s infrastructure delivers:
- Scalable performance in training and inference
- Optimised total cost of ownership with dense GPU setups
- Deployment flexibility in public cloud, dedicated regions, and sovereign cloud environments
Oracle’s use of NVIDIA Blackwell GPUs is a watershed milestone in the growth of cloud-based AI. OCI is well positioned to power AI across industries such as healthcare, banking, manufacturing, and government, thanks to its unparalleled computing performance, integrated networking, and enterprise-ready software.
As the demand for AI capabilities rises, Oracle’s NVIDIA Blackwell-powered infrastructure provides a future-proof basis for creativity. With unrivalled compute, high-speed networking, and full-stack integration, organisations can push the boundaries of artificial intelligence.
