Azure VMware Solution Migration Enterprise Guide (2026)

Azure VMware Solution, native Azure VMs, and Oracle Cloud VMware Solution (OCVS) are the three credible destinations for an on-prem VMware estate that needs to leave the datacenter. AVS wins for Microsoft-anchored enterprises because the SDDC sits inside the Azure governance plane, ExpressRoute lights up native Azure PaaS as a one-hop modernization on-ramp, Microsoft Entra and Azure Policy apply directly, and the Microsoft Customer Agreement vehicle covers the spend. Native Azure VMs wins when the customer has small VMware footprints, long modernization runways, and high modernization appetite — the rebuild cost is real but the result is a cleaner native-PaaS posture. OCVS wins primarily for Oracle-database-anchored enterprises where keeping the database tier physically adjacent to Oracle Exadata Cloud Service is the dominant requirement. EPC Group sizes the decision against the customer workload portfolio in the AVS assessment phase. For Microsoft-strategic enterprises with 200+ VMware-anchored workloads under datacenter pressure, AVS is almost always the right answer.

Host count is the single largest cost lever in AVS, and on-prem VMware over-provisioning rarely transfers cleanly into the AVS density model. EPC Group runs the customer VM portfolio — CPU, memory, storage, IOPS, and growth headroom — against each SKU capacity model. AV36 fits general-purpose enterprise workloads with 576 GB RAM and 36 cores per host. AV52 fits memory-heavy workloads (SQL, SAP, Oracle, large VDI consolidation) with 1.5 TB RAM and 52 cores. AV64 fits high-density VM consolidation with 1 TB RAM and 64 cores. Most enterprise AVS deployments blend SKUs across clusters — AV36 for general production, AV52 for the database tier, AV64 for consolidation density. Realistic host counts for typical enterprise VMware estates land between 12 and 64 hosts across 2 to 6 clusters per region. The assessment-phase capacity model usually recommends 20% to 35% fewer hosts than the customer initial estimate.

Every AVS private cloud lands behind a Microsoft-managed ExpressRoute circuit that terminates inside Azure. ExpressRoute Global Reach peers the AVS ExpressRoute to the customer on-prem ExpressRoute, creating a private Layer-3 path from on-prem datacenter to AVS that bypasses the public internet. NSX-T Tier-0 and Tier-1 gateways inside the AVS SDDC handle the workload-side routing, distributed firewall, and micro-segmentation. Native Azure services — Azure SQL, Key Vault, Blob, Entra ID — reach the AVS workloads through the same ExpressRoute circuit using private endpoint. HCX rides the same path for migration and for ongoing Layer-2 network extension. For tier-1 workloads with multi-AZ resilience requirements, Stretched Cluster on AV36 distributes a vSAN datastore synchronously across two Azure availability zones. The full network design lands in the assessment phase deliverable.

AVS supports a bring-your-own VMware license model for customers who already own Software Assurance-backed VMware entitlements. In practice, the post-Broadcom VMware contract structure has made BYO meaningfully less attractive than the bundled AVS host pricing for most enterprises. Microsoft bundles the VMware Cloud Foundation license inside the AVS host SKU pricing — the customer never holds a direct Broadcom VMware license for the AVS workloads — and AVS host prices have been comparatively stable since the Broadcom transition. Customers with deeply discounted legacy enterprise license agreements should run the math both ways during the assessment, but most newer or smaller customers default to the bundled AVS pricing because the operational simplicity outweighs the residual license-arbitrage opportunity. The cost savings most customers actually capture come from Reserved Instance commitments and from Azure Hybrid Benefit on the guest Windows Server and SQL Server licenses, not from BYO VMware.

AVS and Azure Stack HCI solve different problems despite both being Microsoft-managed hybrid platforms. AVS is a destination service inside an Azure region — the customer leaves the datacenter and runs the VMware workloads inside Azure with Microsoft operating the underlying VMware Cloud Foundation stack. Azure Stack HCI is on-prem hyperconverged infrastructure that runs Hyper-V-based virtualization on customer-owned hardware in the customer datacenter, with Azure as the governance plane. The decision is about where the workload physically lives. Enterprises exiting the datacenter pick AVS. Enterprises that need to keep compute on-prem for latency, data residency, or regulatory reasons pick Azure Stack HCI. Many enterprises run both — AVS as the cloud destination for the bulk of the VMware estate and Azure Stack HCI at the edge sites that cannot leave physical premises. See the Azure Arc Hybrid + Multicloud guide for the broader hybrid model.

HCX Network Extension stretches a Layer-2 segment from the on-prem VMware environment into the AVS SDDC, so workloads migrating from on-prem to AVS keep their IP addresses, MAC addresses, DNS records, and ARP behavior. Without Network Extension every migrated workload would need to be renumbered, every DNS record updated, every firewall rule rewritten, and every application configuration touched — a project-killer at enterprise scale. With Network Extension the migration becomes operationally invisible to the application teams; the workload simply appears in the AVS vCenter with its on-prem IP intact. HCX Mobility Optimized Networking optimizes the post-cutover traffic flow so workloads do not hair-pin through the extension after the migration completes. This is the single biggest reason enterprises pick AVS lift-and-shift over a native Azure VM rebuild — the rebuild requires IP renumbering, AVS does not.

AVS in Azure Commercial holds the standard Azure regulatory certifications including HIPAA, FedRAMP High, SOC 2 Type II, ISO 27001, and PCI DSS. AVS in Azure Government holds the FedRAMP High and DoD IL5 authorizations and is the platform of choice for CMMC 2.0 Level 2 and Level 3 obligated defense contractors. The customer-side compliance work focuses on the workloads inside AVS — applying the NSX-T micro-segmentation, the Defender for Cloud regulatory dashboards, the Microsoft Sentinel detection content, and the Azure Policy compliance initiatives that map to the framework. The AVS BAA covers HIPAA-regulated workloads. EPC Group ships the framework-specific control matrix and the auditor-ready evidence package as part of the AVS Govern phase. The /government-federal-microsoft-consulting-fedramp-cmmc-2026 hub covers the federal-specific story.

AVS itself is billed through the Azure subscription — Reserved Instance host commitments anchor the steady-state model, pay-as-you-go covers migration headroom and burst. Azure Hybrid Benefit on Windows Server and SQL Server typically reduces the gross host bill by 25% to 40%. Annual AVS host spend for typical enterprise deployments ranges from $400K (12-host minimum-viable) to $8M+ (large multi-region estate). The EPC Group AVS Migration Accelerator is a fixed-fee professional services engagement priced between $300K and $1.5M depending on workload count, multi-region scope, regulatory complexity, and the modernization tail the customer wants embedded. Pricing is locked at the assessment-phase deliverable so the customer has board-ready numbers before any cluster gets provisioned. Senior architect on-record from kickoff through go-live, no offshore handoff, no T&M overrun.