Microsoft has opened the public preview of Windows Server 2028, a major platform update designed to let organizations run AI inference closer to the data they need to process. Announced at Microsoft Ignite, the release combines AI infrastructure with updates to security, storage, virtualization and administration.
The goal is to help enterprises place inference workloads where they best fit—whether in the cloud or on infrastructure they control—without building a separate AI stack for every location. The preview includes tools for managing accelerators, setting up inference environments and routing model requests. Microsoft is also working to extend Foundry across local deployments and support open-weight and custom models through Foundry Local, Windows ML, vLLM and Llama.cpp.
Hardware support will be important. AMD’s Instinct MI350P accelerator is expected to enter preview in the first half of 2027 on supported Windows Server systems, including Server 2028. The hybrid AI capabilities are currently limited to customers in an early-access program with specific hardware requirements.
Security, storage and management updates
Windows Server 2028 introduces a hardened Server Core configuration with hundreds of security settings enabled by default. Administrators can loosen settings when workloads require it, but Microsoft says starting from a secure baseline should reduce the time newly deployed servers spend exposed before hardening.
The release also includes isolation features, identity and governance controls for agents, and protections intended to establish trust in virtual machines. Microsoft is developing automatic certificate management and post-quantum cryptography capabilities as part of its longer-term security work.
Native NVMe storage support is enabled by default. Intune management for Windows Servers is available in preview, bringing configuration, compliance and update workflows to eligible servers. Windows Admin Center Virtualization Mode is generally available and lets administrators manage server and virtualization infrastructure together.
Some features are still at an early stage. Microsoft describes Server 2028 as prerelease software and warns that capabilities may change before the final release. Organizations can apply for early access through the Windows Server Technology Adoption Program; NVMe and Intune server management can already be tested.
New AI workloads may require new infrastructure
Running local AI could mean significant spending on accelerators, memory and networking. Existing servers may need Nvidia or AMD GPUs, more RAM and faster connections; 10GbE networking, for example, could become a bottleneck for demanding workloads.
Still, some enterprises may see the release as a practical upgrade even if they are not adopting its AI features immediately. Microsoft’s broader proposition rests on integrating local workloads with Foundry and Azure services, while giving organizations more control over data location, costs and security. The move also reflects interest among some customers in bringing workloads back on premises as public-cloud AI expenses rise.
Compatibility and timing matter
The security and memory defaults could affect applications and tools that worked on Windows Server 2025. Older protocols, legacy authentication and some management scripts may require changes. Microsoft recommends testing applications and operational tooling against the new configurations before deployment, and relaxing protections only when necessary.
There is no immediate need for every organization to upgrade. Windows Server 2025 remains under extended support until 2034, while earlier editions also have years of support remaining. The Long-Term Servicing Channel is designed for organizations that need a stable, feature-locked operating system, including mission-critical and regulated environments.
Some early AI adopters have already built their own Linux clusters or turned to cloud and specialist providers. More cautious organizations may prefer to wait for the technology and its costs to mature, rather than be among the first to navigate the practical challenges of a new platform.
