Rudra Kasturi Inc.

Article · Artificial Intelligence

NVIDIA and Microsoft Expand AI Computing Across Azure and Windows PCs

SUMMARY

NVIDIA and Microsoft announced an expanded collaboration on November 19, covering Azure systems powered by NVIDIA Vera Rubin NVL72, open model customization through Microsoft Foundry, confidential computing and edge AI. The announcement reflects a push to run enterprise AI across cloud data centres and local devices.

NVIDIA and Microsoft are expanding their partnership to help enterprises build and run AI applications across cloud infrastructure, industrial systems and Windows PCs. The November 19 announcement brings together infrastructure, model customization and local AI computing.

The partnership matters because enterprise AI is no longer just about choosing a model. Companies also need computing capacity, secure data handling, deployment options and the ability to run workloads close to where business data is created.

What did NVIDIA and Microsoft announce?

NVIDIA Vera Rubin in Azure: Microsoft is bringing Azure virtual machines powered by NVIDIA Vera Rubin NVL72 systems into preview. These systems are designed for demanding AI reasoning, agent applications, model training and inference.

Open model customization: NVIDIA open models and customization capabilities are becoming available through Microsoft Foundry. This gives enterprise teams another route to adapt models for their own applications and data.

Confidential computing: The companies are working to extend confidential computing protections across bare-metal and rack-scale AI infrastructure. The goal is to protect sensitive models and data while they are being processed.

AI at the edge: The partnership also includes more options for deploying AI closer to industrial systems and Windows PCs, where organisations create and use data.

NVIDIA says Microsoft is also using its technologies to accelerate internal models and services, including NVIDIA Dynamo and RTX PRO 6000 Blackwell GPUs.

Why cloud and local AI both matter

Cloud computing offers scale. Local and edge computing can reduce latency, support disconnected workflows and keep some data closer to its source.

A manufacturer, for example, may use cloud infrastructure to train or customise a model, while running a smaller inference workload on an industrial device. A media company may use cloud AI for large processing jobs and local systems for tasks that need fast response or tighter control over files.

The best architecture depends on cost, data sensitivity, latency and operational requirements. There is no single deployment model that works for every business.

What Indian enterprises should consider

Indian businesses evaluating enterprise AI should ask four questions before committing to infrastructure.

  1. Which workloads genuinely need large cloud accelerators?
  2. Which tasks can run on existing enterprise hardware or edge devices?
  3. What data must remain within specific security or regulatory boundaries?
  4. How will the organisation measure the cost per successful task, not just the cost per token?

The partnership gives Microsoft customers more options, but it does not remove the need for workload testing and cost modelling.

The larger AI infrastructure race

The competition is moving towards full-stack platforms. Cloud providers, chip companies and model developers are working together to make AI easier to train, customise, deploy and operate.

For NVIDIA, the opportunity is to make its accelerated computing platform central to more enterprise workflows. For Microsoft, the partnership strengthens Azure and Foundry as places where customers can build and run AI applications.

Frequently asked questions

What did NVIDIA and Microsoft announce?

They expanded their collaboration across Azure AI infrastructure, open model customization, confidential computing and edge AI.

What is NVIDIA Vera Rubin NVL72?

It is an NVIDIA accelerated computing system that Microsoft is bringing into Azure for demanding AI workloads.

Does this mean AI will run entirely on Windows PCs?

No. The announcement expands options for local and edge AI. Large training and inference workloads will continue to use cloud and data centre infrastructure where appropriate.

Why is confidential computing important for enterprise AI?

It can help protect sensitive data and models while they are being processed, which matters for organisations handling proprietary or regulated information.

Sources