Dell and Microsoft team up to bring Nvidia HGX B300 to private clouds
The partnership allows companies to run AI models on their own hardware to meet strict data privacy and sovereignty requirements.

The short version
- The integration combines Dell hardware, Nvidia HGX B300 chips, and Microsoft's Azure Local and Foundry Local software.
- 85% of enterprises say their AI workloads require a mix of both on-premises and public cloud infrastructure.
- Sovereignty and latency influenced 51% of respondents in a recent IDC survey, ahead of cost concerns.
- The setup allows for large-scale, multi-GPU inference and agentic workloads in customer-controlled environments.
Why it matters
For businesses in highly regulated sectors like healthcare or finance, this provides a way to use advanced AI without moving sensitive data to a public cloud. It balances the need for high-performance computing with the necessity of maintaining strict data residency controls.
The push for private AI
Dell and Microsoft are broadening their collaboration on private cloud systems to accommodate enterprise-level AI tasks. The new initiative centers on the Nvidia HGX B300 accelerated computing platform, which is built to handle demanding AI processes. This move specifically targets organizations that want to run inference—the process of using a trained AI model to generate outputs—closer to their own sensitive data.
By combining Dell hardware with Microsoft's Azure Local and Foundry Local software, the companies are offering a hybrid approach. This allows businesses to deploy AI outside of standard public cloud environments while keeping a tight grip on their data processing. It is not a move away from the cloud, but rather a way to balance different operational needs.
According to a survey by IDC, 85% of enterprises reported that their AI workloads require both on-premises and public cloud infrastructure. This trend suggests that companies are looking for a middle ground where they can utilize the scale of the cloud while maintaining the security of local hardware. The goal is to create a flexible environment that meets various corporate requirements.
How the hardware integrates
The architecture brings Nvidia HGX B300 systems into Microsoft's Azure Local environment through validated hardware partners like Dell. Azure Local serves as the platform for running workloads in environments that the customer actually controls. This setup is designed to support large-scale, multi-GPU inference and agentic workloads, which are AI systems capable of performing complex tasks autonomously.
Foundry Local is the specific software component that enables organizations to run these AI models locally. By keeping the inference close to the enterprise data, companies can reduce the risks associated with moving information across networks. Microsoft confirmed that Foundry Local is generally available on Azure Local and supports the expanded Nvidia hardware options.
Dell, along with HPE and Lenovo, are the primary hardware partners for these validated systems. While Microsoft has not provided specific delivery dates for these configurations, the integration is intended to simplify the layers of infrastructure. It aims to streamline lifecycle management, firmware coordination, and general support for the hardware and software stack.
Who is affected by this shift
Regulated organizations are the primary audience for this type of private cloud expansion. Financial services, healthcare providers, and government agencies often face strict legal requirements regarding where their data can reside. For these groups, the ability to control workload placement is a major factor in their infrastructure decisions.
IDC expects sovereign private clouds to be among the fastest-growing segments of private cloud through 2028. This growth is driven by concerns over data sovereignty and the need for lower latency. Organizations are increasingly looking for ways to satisfy these requirements without sacrificing the performance of modern AI models.
However, moving AI workloads to local infrastructure introduces a significant operational burden for IT teams. Companies become responsible for managing the GPU capacity, storage, networking, power, and cooling required to run the systems. While the Dell and Microsoft integration simplifies some of these layers, IT teams still must manage the underlying capacity and optimize the workloads themselves.
The economics of local AI
The financial logic for running AI on owned infrastructure depends heavily on how much the system is used. IDC notes that hosting AI locally is financially viable when the inference tasks are high-volume, consistent, and easy to predict. For large and frontier models running sustained workloads, the total cost of ownership can be significantly lower than using cloud infrastructure as a service once facility costs are included.
The principle is familiar to many IT departments: organizations that can accurately forecast their capacity requirements can often run workloads more cheaply on their own hardware. The challenge currently lies in the fact that many enterprises are still figuring out which specific AI workloads they will run and how much computing power those applications will actually need.
Analysts caution that data sovereignty should not be viewed as the only reason to move to on-premises systems. While public cloud providers offer many data controls, the primary commercial advantage of a specific environment is often the most important consideration. Companies must weigh the costs of managing their own hardware against the convenience and flexibility of a public cloud provider.
What to watch
- Specific delivery dates for the Dell-validated Nvidia HGX B300 configurations.
- How many enterprises successfully transition to private AI inference versus staying on public clouds.
- The development of more open-weight models that might simplify the deployment choices for local infrastructure.
Try it
Check if your organization's data residency requirements allow for public cloud AI, or if you need to look into sovereign private cloud options like Azure Local.
Sources and links
- ReportingTechTarget: the article we sourced
Read the original at TechTarget
First reported by TechTarget. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.