Microsoft combines accelerated computing with cloud scale engineering to convey superior AI capabilities to our clients. For years, we’ve labored with NVIDIA to combine {hardware}, software program and infrastructure to energy lots of immediately’s most vital AI breakthroughs.
What’s new at NVIDIA GTC
- Expanded Microsoft Foundry capabilities to construct, deploy and function production-ready AI brokers on NVIDIA accelerators and open NVIDIA Nemotron fashions
- New Azure AI infrastructure optimized for inference-heavy, reasoning-based workloads, together with the primary hyperscale cloud to energy on next-generation NVIDIA Vera Rubin NVL72 methods
- Deeper integration throughout Microsoft Foundry, Microsoft Material and NVIDIA Omniverse libraries and open frameworks to help Bodily AI methods from simulation to actual‑world operations
From Frontier fashions to production-ready brokers
On the basis of this method is Microsoft Foundry: serving because the working system for constructing, deploying and working AI at enterprise scale. Foundry builds on Azure to convey collectively fashions, instruments, knowledge and observability right into a single system designed for manufacturing brokers. At the moment we’re increasing these capabilities throughout Foundry Agent Service and NVIDIA Nemotron fashions.
The following-generation Foundry Agent Service and Observability in Foundry Management Aircraft at the moment are typically accessible, enabling organizations to construct and function AI brokers at manufacturing scale. Foundry Agent Service permits groups to rapidly develop brokers that motive, plan and act throughout instruments, knowledge and workflows. As soon as created, Foundry Management Aircraft gives the developer end-to-end visibility into agent conduct, unlocking each developer productiveness in addition to enterprise belief. Firms corresponding to Corvus Power are already utilizing Foundry to interchange guide inspection workflows with agent-driven operational intelligence throughout their international fleet.
We’re additional simplifying the trail from prototype to manufacturing with the provision of Voice Stay API integration with Foundry Agent Service, in public preview, which permits builders to construct voice-first, multimodal, real-time agentic experiences. This pairs with the overall availability of a refreshed Microsoft Foundry portal and expanded integrations for Palo Alto Networks’ Prisma AIRS and Zenity, delivering deeper builder experiences and runtime safety throughout all the agent lifecycle.
NVIDIA Nemotron fashions are additionally now accessible via Microsoft Foundry, becoming a member of the widest number of fashions on any cloud, together with the newest reasoning, frontier and open fashions. This bolsters our latest partnership announcement bringing Fireworks AI to Microsoft Foundry, enabling clients to fine-tune open-weight fashions like NVIDIA Nemotron into low-latency belongings that may be distributed to the sting.
Scaling AI infrastructure for the world’s most demanding workloads
Inference AI workloads are reshaping value, efficiency and system design necessities. To operationalize agentic AI at scale, clients want purpose-built infrastructure for inference‑heavy, reasoning‑primarily based workloads that may be deployed and operated constantly throughout international and controlled environments.
Microsoft’s AI infrastructure method is engineered to seamlessly convey next-generation NVIDIA methods into Azure datacenters which can be designed for energy, cooling networking and fast generational upgrades. This enables our clients to maneuver with pace and agility and keep at the vanguard from era to era.
In lower than a 12 months, we’ve deployed lots of of 1000’s of liquid-cooled Grace Blackwell GPUs throughout our international datacenter footprint, and now we’re excited to be the first hyperscale cloud to energy on NVIDIA’s latest Vera Rubin NVL72 in our labs. Over the subsequent few months, Vera Rubin NVL72 can be rolled out into our trendy, liquid-cooled Azure datacenters.
Microsoft’s infrastructure innovation with NVIDIA additionally extends to sovereign and controlled environments to provide clients management of each the place AI runs and the way it evolves over time. Just lately, we introduced Foundry Native help for contemporary infrastructure and enormous AI fashions, and immediately we now have preliminary help for NVIDIA Vera Rubin platform on Azure Native, extending accelerated AI capabilities to customer-controlled environments. This method permits organizations to plan for next-generation AI workloads, together with reasoning-based and agentic methods, whereas sustaining Azure-consistent operations, governance and safety via our unified software program layer with Azure Arc and Foundry Native.
Bringing AI into the bodily world
As AI strikes past digital experiences, Microsoft and NVIDIA are collaborating to help the subsequent wave of Bodily AI. At GTC, this work facilities on NVIDIA Bodily AI Knowledge Manufacturing unit Blueprint, with Microsoft Foundry because the platform for internet hosting and working Bodily AI methods on Azure at cloud scale.
By integrating this blueprint with Azure providers as a part of a Bodily AI Toolchain, Microsoft permits builders to construct, practice and function bodily AI and robotics workflows that join bodily belongings, simulation and cloud coaching environments into repeatable, enterprise-grade pipelines. To help, we’re introducing a public Azure Bodily AI Toolchain GitHub repository built-in with the Nvidia Bodily AI Knowledge Manufacturing unit and with core Azure providers.
To additional the impression of AI in actual‑world, bodily environments, immediately Microsoft and NVIDIA are deepening the combination between Microsoft Material and NVIDIA Omniverse libraries, connecting reside operational knowledge with bodily correct digital twins and simulation. This enables organizations to see what’s occurring throughout their bodily methods, perceive it in actual time and use AI to determine what to do subsequent. In apply, clients in manufacturing and operations and past are utilizing this method to maneuver past dashboards and alerts to coordinated, AI‑pushed motion throughout machines, amenities and workflows.
From innovation to impression
Microsoft is delivering dependable, manufacturing‑scale AI by bringing collectively its international AI infrastructure, platforms and actual‑world methods with the newest innovation from NVIDIA. For purchasers, this implies the flexibility to function intelligence repeatedly, working inference-heavy, reasoning-based and bodily AI workloads with the efficiency, safety and governance required for actual companies and controlled industries.
Whether or not powering always-on brokers, scaling next-generation AI infrastructure or deploying clever methods in factories, vitality amenities and sovereign environments, Microsoft and Nvidia are serving to clients transfer quicker from perception to motion.
Yina Arenas leads product technique and execution for Microsoft Foundry, overseeing the tip–to–finish AI product portfolio, infrastructure, developer experiences and basis mannequin integration throughout OpenAI, Anthropic, Mistral, DeepSeek and others. She delivers an enterprise prepared, manufacturing grade AI platform trusted by international clients for safe, dependable and scalable AI.