Hyper-V on Windows Server 2025: New Features and What’s Changed

Hyper-V on Windows Server 2025 brought the biggest scalability jump the platform has seen in years, and Microsoft has kept building on it through 2026. If you’re planning a Hyper-V deployment or upgrade, here’s what’s actually new and what it means in practice. Quick Facts Generation 2 VMs now support up to 2,048 virtual CPUs […]

Hyper-V on Windows Server 2025 new features illustration

Hyper-V on Windows Server 2025 brought the biggest scalability jump the platform has seen in years, and Microsoft has kept building on it through 2026. If you’re planning a Hyper-V deployment or upgrade, here’s what’s actually new and what it means in practice.

Quick Facts

  • Generation 2 VMs now support up to 2,048 virtual CPUs and 240 TB of memory per VM
  • Hosts scale to 4 petabytes of memory and 2,048 logical processors
  • Maximum VHDX virtual disk size remains 64 TB
  • GPU partitioning allows shared GPU access across VMs, with live migration support
  • Workgroup clusters enable AD-free Hyper-V clustering, aimed at edge deployments
  • A free Windows Admin Center conversion wizard migrates VMware VMs to Hyper-V via agentless online replication

Scalability: How Much Bigger?

The headline number is per-VM scale: a Generation 2 VM can now be configured with up to 2,048 virtual CPUs and 240 TB of memory, which covers essentially any workload short of the largest in-memory databases. Host-level limits moved up in step — up to 4 petabytes of memory and 2,048 logical processors per host — so the ceiling on host consolidation ratios has moved dramatically higher too. Virtual disk size is unchanged at 64 TB per VHDX, which is still the practical limit to plan storage layouts around.

In practice, most organisations won’t get near these ceilings, but they matter for consolidation planning — fewer, larger hosts become a more realistic strategy than they were on Windows Server 2022, provided your storage and networking can keep up.

GPU Partitioning

GPU partitioning lets multiple VMs share a single physical GPU, with each VM getting a defined slice of GPU resources — and critically, it now supports live migration, so a VM using a GPU partition can move between hosts without losing its GPU allocation. This matters for AI/ML inferencing workloads, VDI with graphics acceleration, and any scenario where you previously had to choose between GPU passthrough (fast, but blocks live migration) and no GPU access at all.

Workgroup Clusters

Workgroup clusters are Windows Server Failover Clusters where the Hyper-V nodes aren’t domain-joined, with live migration still supported between nodes. This is aimed squarely at edge and branch-office deployments where standing up (or maintaining connectivity to) a domain controller isn’t practical — a small remote site can now run a resilient two- or three-node Hyper-V cluster without any Active Directory dependency.

Migrating from Older Hosts

If you’re moving from Windows Server 2019/2022 Hyper-V hosts, VM configuration version upgrades and cluster rolling upgrades work the same way they always have — bring nodes into the new cluster gradually and update VM configuration versions once all nodes are upgraded, not before. If you’re migrating from VMware instead, see the dedicated guide below on the free agentless conversion wizard.

For the full base setup and configuration walkthrough, see our Hyper-V on Windows Server 2025 setup guide.


Discover more from TechyGeeksHome

Subscribe to get the latest posts sent to your email.

Andrew Armstrong

Andrew Armstrong is a UK-based IT professional with 26+ years of hands-on experience in Windows, Windows Server, SCCM/ConfigMgr, Active Directory, PowerShell, and enterprise infrastructure.

He founded TechyGeeksHome in 2010 and has published over 1,500 practical guides covering real-world IT problems and solutions. When not solving IT problems,

Andrew develops free Windows utilities including Ultimate Settings Panel, which has been downloaded over 850,000 times.

Leave a Reply

Your email address will not be published. Required fields are marked *