NVIDIA and Microsoft are deepening their collaboration to drive advancements in agentic Artificial Intelligence, extending from cloud computing infrastructure to personal computers. At the Microsoft Build event, Microsoft introduced Microsoft Discovery, a platform designed to empower researchers and revolutionize the research and development (R&D) pipeline through agentic Artificial Intelligence tools. The platform aims to streamline and expedite discovery processes across industries, promising faster time-to-market for new products and scientific breakthroughs.
Microsoft Discovery incorporates NVIDIA’s specialized microservices, including ALCHEMI NIM, which enhances AI-based chemical simulations for materials science, and BioNeMo NIM for drug discovery workflows. These integrations provide researchers with accelerated data analysis, property prediction, and candidate recommendation, leading to meaningful gains in research speed and accuracy. In real-world testing, Microsoft researchers using the platform discovered a new coolant prototype for data centers in just under 200 hours, a process that previously required months or years.
On the infrastructure front, Microsoft is deploying tens of thousands of NVIDIA GB200 NVL72 rack-scale systems across its Azure data centers. These state-of-the-art machines offer significant performance improvements, including up to 35 times more inference throughput compared to previous Azure models, leveraging up to 72 NVIDIA Blackwell GPUs per rack and advanced liquid cooling technologies. Custom server architectures, NVLink interconnects, and NVIDIA Quantum InfiniBand networking deliver seamless scalability for large-scale generative and agentic Artificial Intelligence workloads, ensuring both higher efficiency and lower costs for enterprises.
The partnership further extends to healthcare and reasoning workloads through the integration of NVIDIA Llama Nemotron models and BioNeMo NIM microservices within Azure AI Foundry. These tools provide optimized, enterprise-grade inferencing for complex decision-making, digital biology, drug discovery, and medical imaging applications. For example, services like ProteinMPNN, RFDiffusion, and OpenFold2 accelerate protein science and molecular modeling, vastly improving patient care and innovation timelines. Organizations can now deploy these high-performance Artificial Intelligence agents rapidly and at scale, benefiting from secure, containerized platforms tailored to domain needs.
Finally, NVIDIA and Microsoft are bringing generative Artificial Intelligence capabilities directly to end-users by optimizing the AI inferencing stack for Windows 11 and RTX AI PCs. With an updated TensorRT package for RTX, developers gain 8x smaller deployment sizes and just-in-time, on-device engine building, streamlining application development and boosting performance on over 100 million RTX-enabled devices. TensorRT integration with Windows ML enables broad hardware compatibility and out-of-the-box state-of-the-art performance. These developments pave the way for a new era of intelligent applications in research, enterprise, and on consumer devices.
