Microsoft is positioning Windows Server 2028 as a platform for running AI workloads alongside traditional enterprise applications. For organizations, the release introduces new options for hybrid AI deployment, but it also creates decisions around infrastructure, governance, security, and data residency.
Key Takeaways:
As enterprises evaluate where AI workloads should run, many are discovering that the cloud is only part of the solution. Data residency requirements, latency concerns, regulatory obligations, and the need to process sensitive information close to business systems are driving demand for on-premises and hybrid AI infrastructure. Moreover, organizations must modernize aging server environments while strengthening security and simplifying management across increasingly distributed estates.
To address these challenges, Microsoft announced the public preview of Windows Server 2028, which positions the release as a platform for “enterprise hybrid intelligence.” This update combines new AI infrastructure capabilities with security, storage, virtualization, and management enhancements designed to help organizations run both traditional workloads and AI-powered services across on-premises, edge, and cloud environments.
Enterprise IT teams face the biggest question of where AI inference should run. Cloud-based AI services remain attractive for many scenarios, but organizations increasingly want the flexibility to run AI close to their applications, operational systems, and sensitive data. This is particularly important for workloads affected by latency requirements, intellectual property concerns, regulatory restrictions, or data sovereignty mandates.
The challenge is that many organizations do not want to build separate infrastructure stacks for traditional workloads and AI systems. Enterprises are looking for ways to deploy AI capabilities on the same platforms already used to run databases, virtual machines, file services, and line-of-business applications.
Microsoft has introduced a new hybrid AI strategy for Windows Server 2028. The company says it is making long-term investments to extend Microsoft Foundry to enterprise infrastructure, allowing organizations to run AI inference on local servers while remaining connected to broader hybrid AI architectures.
At the hardware layer, Microsoft announced support for new AI acceleration technologies, including the preview of AMD Instinct MI350P accelerators for supported Windows Server environments beginning in the first half of 2027. At the platform level, Microsoft is developing tools to deploy inference environments, manage accelerated hardware, and route AI workloads to appropriate execution environments.
Microsoft also plans to support multiple AI runtime technologies, including Foundry Local, Windows ML, vLLM, and Llama.cpp, giving organizations greater flexibility when deploying open-weight and custom AI models.
To support governance requirements, Microsoft is integrating broader AI management technologies into the platform, including Microsoft eXecution Containers for workload isolation, Microsoft Foundry model catalogs, and Agent 365 for AI agent identity and governance. These capabilities are intended to help enterprises run AI systems close to their applications and data while maintaining control over security and compliance.
Microsoft is also using this release to strengthen the server platform’s security foundation. The company says Windows Server 2028 will include a hardened Server Core configuration by default and a new security baseline designed to reduce the effort required to establish secure deployments
In Windows Server 2008, other security enhancements include Trusted Launch protections for virtual machines, which help organizations verify boot integrity across their VM estates, and early support for Confidential VMs, which are designed to provide integrity-protected runtimes on supported hardware.
Microsoft is also introducing automatic certificate management capabilities and expanding support for post-quantum cryptography, reflecting growing industry concerns about the long-term risks quantum computing could pose to current encryption standards.
Microsoft is also targeting organizations running performance-sensitive workloads. Windows Server 2028 introduces native NVMe storage support enabled by default, along with NVMe over TCP and RDMA transport support. The company is also adding the ability to boot directly from ReFS (Microsoft’s resilient file system).
Microsoft says these enhancements are intended to help customers take greater advantage of modern storage hardware while improving scalability, throughput, and performance for demanding enterprise applications.
According to Microsoft, managing large server estates remains a challenge for many organizations, particularly those operating across datacenters, remote offices, and cloud environments. Microsoft says customers increasingly want consistent management tools for both clients and servers.
To address that need, Microsoft announced that Microsoft Intune management for Windows Server is now available in preview. This feature extends existing endpoint-management capabilities (including configuration management, compliance policies, application deployment, and update controls) to eligible persistent Windows Server environments.
Microsoft also highlighted the continued role of Azure Arc for cloud-native server management and announced general availability of Windows Admin Center Virtualization Mode (WAC vMode), which provides administrators with deeper visibility into virtualization hosts and server infrastructure.
For enterprise IT leaders, Windows Server 2028 represents more than a routine server refresh. Microsoft’s strategy reflects growing demand for hybrid AI infrastructure that allows organizations to run AI workloads in the cloud, at the edge, or within corporate datacenters. Support for AI accelerators, multiple inference runtimes, and AI governance tools gives enterprises more options for deploying AI without adopting entirely new infrastructure platforms.
At the same time, the new capabilities introduce additional complexity. Organizations will need to determine where AI workloads should run based on factors such as cost, latency, regulatory requirements, security controls, and data residency obligations. Running AI on-premises may offer greater control, but it also requires investments in hardware, management, governance, and operational expertise.
As for the security and management improvements, the new features such as hardened Server Core deployments, Trusted Launch, Confidential VM support, automatic certificate management, post-quantum cryptography preparations, and Intune-based server management may help organizations strengthen security and reduce administrative overhead. However, enterprises will still need to evaluate compatibility, update operational processes, and prepare for long-term cryptographic modernization efforts.