NVIDIA and Microsoft Accelerate Agentic AI Innovation, From Cloud to PC

Agentic AI is redefining scientific discovery and unlocking breakthroughs across industries. Through deepened collaboration, NVIDIA and Microsoft are delivering innovations that accelerate agentic AI-powered applications — from the cloud to the PC.

At Microsoft Build, Microsoft unveiled Microsoft Discovery, an extensible platform empowering researchers to transform the discovery process with agentic AI. This platform will help accelerate product development and streamline R&D workflows across industries.

Microsoft Discovery will integrate:

NVIDIA ALCHEMI NIM, which optimizes AI inference for chemical simulations — ideal for materials science
NVIDIA BioNeMo NIM, which accelerates drug discovery using pretrained AI workflows

In internal testing, Microsoft researchers used the platform to discover a novel coolant prototype for data centers in under 200 hours — a task that traditionally takes months or even years.

Deploying NVIDIA GB200 Systems at Scale

Microsoft is deploying hundreds of thousands of NVIDIA Blackwell GPUs via GB200 NVL72 rack-scale systems in AI-optimized Azure data centers worldwide.

Key benefits:

10x performance over today’s fastest supercomputer
100% renewable energy powering Azure AI data centers by end of 2025
Up to 35x inference throughput with ND GB200 v6 VMs, compared to previous-generation ND H100 v5 instances

The performance gains are enabled by:

Custom server designs
High-speed NVIDIA NVLink
NVIDIA Quantum InfiniBand networking

These innovations support massive generative and agentic AI workloads with greater efficiency and scalability.

Software Optimizations Driving Next-Level AI Performance

Microsoft and NVIDIA are compounding performance gains with continuous software optimizations across all NVIDIA GPU architectures on Azure.

This approach results in:

Higher developer productivity
Lower total cost of ownership
Increased workload efficiency per dollar and per watt

Expanded NIM Microservices in Azure AI Foundry

NVIDIA and Microsoft are also expanding Azure AI Foundry with new NIM microservices and open reasoning models.

Now available:

NVIDIA Llama Nemotron Super and Nano models — offer advanced multistep reasoning, coding, and agentic capabilities with 20% higher accuracy and 5x faster inference
Healthcare-focused BioNeMo NIM microservices, including:
- ProteinMPNN
- RFDiffusion
- OpenFold2

These services enable high-performance inferencing for digital biology, molecular modeling, and genomics — accelerating scientific and medical innovation.

Generative AI Comes to Windows 11 with RTX AI PCs

NVIDIA and Microsoft are bringing generative AI capabilities to the desktop with RTX AI PCs and a new AI inferencing stack for Windows 11.

Highlights:

TensorRT for RTX combines JIT, on-device engine building, and 8x smaller packages for seamless AI deployment
Native support in Windows ML, enabling cross-hardware compatibility and optimized performance for AI apps

TensorRT for RTX is now available in the Windows ML preview and will be released as a standalone SDK via NVIDIA Developer in June.

Conclusion

From enterprise data centers to everyday PCs, NVIDIA and Microsoft are building the foundation for a new era of agentic and generative AI. These advancements will redefine how developers, researchers, and creators build the future — faster, smarter, and more sustainably.