How AVS pricing actually works after the October 2025 licensing change, what the two-vendor cost structure looks like, when AVS is the right migration path, and where the economics break down. This guide explains every cost layer so enterprises can model the real price of AVS before committing.
Part of the Broadcom/VMware Advisory series
📚 Broadcom/VMware Knowledge HubFor the full Broadcom/VMware licensing reference including contract strategy, alternatives, and cost optimisation, start there.
Azure VMware Solution is a first-party Microsoft Azure service that provides dedicated bare-metal hosts running the full VMware Cloud Foundation stack — ESXi, vCenter Server, vSAN, NSX, and optionally HCX for migration — in Azure datacentres. Microsoft manages the physical infrastructure (servers, networking, storage), performs ESXi patching and upgrades, and provides unified support through Azure support channels. You manage your vSphere environment through vCenter and NSX Manager just as you would on-premises.
The critical distinction: AVS is not a re-platforming exercise. Your existing VMware VMs migrate to AVS with minimal or no modification. The same tools work. The same skills apply. The same third-party integrations (Veeam, Commvault, CrowdStrike) function on AVS as they do on-premises. This is what makes AVS attractive for enterprises that want to exit their data centres but cannot invest the time or risk required to migrate workloads to Azure-native IaaS (Virtual Machines) or PaaS services.
However, AVS is not Azure-native. Your workloads run on VMware inside Azure, which means you are still a VMware customer, still subject to Broadcom’s licensing terms, and still exposed to Broadcom’s pricing trajectory. AVS solves the infrastructure problem (you no longer own and maintain physical servers) but does not solve the VMware licensing problem. Understanding this distinction is essential for a clear-eyed cost evaluation.
Before October 2025, AVS was sold as a single line item: you paid Microsoft a per-node hourly rate that included both the Azure infrastructure and the VMware VCF software licence. Broadcom licensed VMware to Microsoft under a wholesale arrangement, and Microsoft passed the cost through as part of the AVS node price. Simple, predictable, one vendor.
That model ended on October 15, 2025. Broadcom announced that all hyperscaler VMware offerings — AVS on Azure, Elastic VMware Service on AWS, and Google Cloud VMware Engine — would shift to a bring-your-own-licence (BYOL) model. The implications for AVS customers are significant.
New nodes purchased after October 15, 2025 require the customer to provide a portable VMware Cloud Foundation subscription purchased directly from Broadcom (or a Broadcom channel partner). Microsoft’s AVS node price now covers only the Azure infrastructure and managed service — the VMware software cost is separate and billed by Broadcom.
Existing Reserved Instances purchased before October 15, 2025 are honoured through the end of their term. Customers who locked in 3-year or 5-year RIs with VCF included before the cutoff can continue operating without any change until their RI expires. This creates a window of protection for forward-planning customers who acted early.
Existing pay-as-you-go deployments with VCF included can continue operating without change through October 31, 2026. After that date, PAYG customers must provide a VCF licence key from Broadcom to continue using AVS.
The net result: AVS is now a two-vendor cost structure. You negotiate and pay Microsoft for the compute infrastructure. You negotiate and pay Broadcom for the VMware software. The simplicity that made AVS attractive — one bill, one vendor, one support channel — has been materially eroded.
Microsoft charges for AVS on a per-node basis. Each node is a dedicated bare-metal server in an Azure datacentre. Pricing varies by region and node type.
| Node Type | Cores | RAM | Storage (Raw) | Approx. PAYG (US East) | Approx. 3-Year RI (US East) |
|---|---|---|---|---|---|
| AV36P | 36 cores (dual 18-core Intel) | 768 GB | ~19 TB NVMe | ~$6.50–$8.00/hr per node | ~$3.50–$4.50/hr per node |
| AV52 | 52 cores (dual 26-core Intel) | 1,536 GB | ~38 TB NVMe | ~$9.00–$11.00/hr per node | ~$5.00–$6.50/hr per node |
| AV64 | 64 cores | 1,024 GB | ~30 TB NVMe | ~$8.50–$10.50/hr per node | ~$4.50–$6.00/hr per node |
Key pricing notes: These are BYOL prices (infrastructure only, no VCF included). The VCF-included prices were approximately 20–35% higher before the October 2025 cutoff. Reserved Instances provide 30–50% savings over PAYG for 1, 3, or 5-year terms and must be purchased through an Azure Enterprise Agreement. A minimum of 3 nodes is required per private cloud cluster, so the minimum PAYG cost (3 × AV36P) is approximately $14,000–$17,500/month in Azure infrastructure alone. Pricing varies significantly by region — APAC and European regions are typically 15–30% more expensive than US East.
The portable VCF subscription is purchased directly from Broadcom or a Broadcom channel partner. It is licensed per physical core with the same 16-core minimum per CPU that applies to on-premises VCF licensing.
| AVS Node Type | Physical Cores per Node | Minimum 3-Node Cluster Cores | Estimated VCF Annual Cost (Negotiated) |
|---|---|---|---|
| AV36P | 36 | 108 | $23,760–$37,800 108 cores × $220–$350/core/yr |
| AV52 | 52 | 156 | $34,320–$54,600 156 cores × $220–$350/core/yr |
| AV64 | 64 | 192 | $42,240–$67,200 192 cores × $220–$350/core/yr |
The VCF subscription includes vSphere, vSAN, NSX (base), and Aria Operations. Advanced NSX features (vDefend Firewall with Advanced Threat Prevention) are separate add-ons purchased from Broadcom. HCX for migration is available but may require a separate licence depending on your Broadcom agreement terms. The VCF licence is portable — meaning the same licence can cover on-premises hosts and AVS nodes, as long as total registered cores across all environments do not exceed your Broadcom entitlement.
Combining both layers into worked scenarios reveals the true all-in cost of AVS. The following models use mid-range negotiated pricing for both Microsoft and Broadcom components.
| Deployment Size | Nodes | Total Cores | Azure Infra (Annual, 3yr RI) | Broadcom VCF (Annual) | Total Annual Cost | Total 3-Year Cost |
|---|---|---|---|---|---|---|
| Small (minimum cluster) | 3 × AV36P | 108 | $110,000 | $29,000 | $139,000 | $417,000 |
| Mid-market | 10 × AV36P | 360 | $365,000 | $90,000 | $455,000 | $1,365,000 |
| Enterprise | 30 × AV52 | 1,560 | $1,450,000 | $375,000 | $1,825,000 | $5,475,000 |
The Broadcom VCF subscription represents approximately 17–21% of the total AVS cost in these scenarios. The Azure infrastructure is the dominant cost at 79–83%. This ratio is important for negotiation strategy: aggressive Broadcom negotiation reduces a meaningful but minority portion of the total bill, while optimising the Azure component (node right-sizing, Reserved Instances, storage efficiency) has a larger absolute impact.
Data leaving Azure (egress) is charged at standard Azure rates: approximately $0.087/GB for the first 10 TB/month, with volume discounts at higher tiers. For organisations with significant data transfer between AVS and on-premises systems (backup replication, database synchronisation, user-facing applications), egress charges can add $500–$5,000+/month depending on volume. This cost does not exist for on-premises VMware deployments and is frequently underestimated in AVS business cases.
Each AVS node includes local NVMe storage consumed by vSAN. If your workloads require more storage than the local NVMe provides, you must attach external storage: Azure NetApp Files, Azure Elastic SAN, or third-party solutions like Pure Cloud Block Store. External storage adds $0.10–$0.30/GB/month depending on tier and performance requirements. For storage-heavy workloads, this cost can be substantial and is the primary reason some AVS deployments end up larger (more nodes) than initially planned — because adding nodes is the only way to add vSAN storage without external solutions.
AVS requires network connectivity to your on-premises environment and other Azure services. ExpressRoute (dedicated private connection) is the recommended option for production deployments, costing $200–$3,000+/month depending on bandwidth and peering location. VPN Gateway is a lower-cost alternative ($150–$500/month) but with less predictable performance. These are ongoing monthly costs that persist for the life of the AVS deployment.
VMware HCX is the standard tool for migrating VMs from on-premises to AVS with minimal downtime. HCX Advanced is included with AVS. HCX Enterprise (which adds features like Replication Assisted vMotion for large-scale migrations) may require a separate licence from Broadcom depending on your VCF entitlement. Professional services for migration planning and execution typically run $30,000–$150,000 depending on the number of VMs and complexity.
The comparison between AVS and on-premises VMware is not purely a licensing question — it is an infrastructure economics question. AVS eliminates the capital expenditure of purchasing servers, the operational cost of running a data centre, and the staffing cost of managing physical infrastructure. But it introduces a continuous operational expense that may exceed on-premises TCO for steady-state workloads.
| Factor | AVS | On-Premises VMware | Advantage |
|---|---|---|---|
| Capital expenditure | $0 (cloud opex) | $500K–$2M+ (servers, storage, networking) | AVS |
| Annual licensing (VMware) | Same VCF cost (BYOL) | Same VCF cost | Neutral (post-October 2025) |
| Infrastructure management | Microsoft-managed (patching, hardware) | Customer-managed (team of 2–5+) | AVS |
| Data centre costs | $0 | $50K–$300K+/year (power, cooling, space) | AVS |
| Flexibility to scale | Add/remove nodes in hours | Procure and install hardware (weeks/months) | AVS |
| Data egress | $500–$5,000+/month | $0 (all traffic on-premises) | On-premises |
| Performance predictability | Shared Azure fabric (high, but not dedicated network) | Dedicated infrastructure (full control) | On-premises |
| 5-year steady-state TCO | Often 10–30% higher than on-prem | Lower for stable, predictable workloads | On-premises (at scale) |
| Data centre exit timeline | Immediate (no hardware to decommission) | N/A (you are the data centre) | AVS |
AVS wins when the primary driver is data centre exit (lease expiration, consolidation, disaster recovery), when capital expenditure constraints prevent hardware refresh, when the organisation lacks infrastructure staff and wants Microsoft to manage the physical layer, or when workloads are temporary or highly variable (seasonal peaks, M&A integration, project-based environments that will be decommissioned).
On-premises wins when the workload is stable and predictable over a 5-year horizon, when the organisation already owns current-generation hardware with remaining useful life, when data sovereignty requirements prevent cloud placement, or when data egress volumes would make cloud hosting disproportionately expensive.
AVS is often positioned as a stepping stone to Azure-native: move to AVS first (lift-and-shift), then gradually re-platform to Azure VMs and PaaS services. This is a legitimate strategy, but the economics of the “two-hop” approach should be evaluated honestly.
If your workloads are primarily Windows/Linux VMs with standard configurations, migrating directly to Azure VMs (IaaS) using Azure Migrate avoids the AVS intermediate step entirely. Azure VMs do not require VMware licensing. The per-VM cost is typically 40–60% lower than the equivalent AVS node cost for compute-bound workloads. The trade-off: direct migration requires more planning (OS compatibility, driver changes, network reconfiguration) and may involve application-level modifications. For organisations with the technical capacity to invest in proper migration engineering, going direct to Azure-native is almost always cheaper in the long run.
If your environment has hundreds or thousands of VMs with complex interdependencies, VMware-specific configurations (vGPU, NSX-dependent networking, vSAN-dependent storage policies), or tight migration timelines (data centre lease expiring), AVS provides a low-risk landing zone. You migrate everything to AVS quickly using HCX, then re-platform to Azure-native over 12–36 months as time and resources allow. The risk: many organisations get stuck on AVS permanently because the re-platforming effort is deprioritised once workloads are running. This creates a “cloud VMware tax” where you pay cloud infrastructure rates plus VMware licensing indefinitely.
A 30-node AVS deployment at $1.8M/year that takes 3 years to fully re-platform to Azure-native costs $5.4M in AVS fees before the re-platforming is complete. If the Azure-native target state costs $900K/year, the 5-year total (3 years AVS + 2 years native) is $7.2M. A direct-to-native migration that costs $500K more in professional services but is completed in Year 1 produces a 5-year total of $4.1M ($500K migration + $3.6M in Azure-native fees). The two-hop approach costs $3.1M more over 5 years in this scenario. The AVS stepping stone is only economically justified when the migration timeline or technical complexity makes direct-to-native impossible in the required timeframe.
Optimise the Microsoft component first. The Azure infrastructure is 80% of the AVS bill. Reserved Instances (1, 3, or 5-year) provide 30–50% savings over PAYG. Commit to RIs for your baseline node count and use PAYG only for burst capacity. If you have an Azure Enterprise Agreement, negotiate AVS-specific pricing as part of your overall Azure commitment.
Right-size your node selection. AV36P, AV52, and AV64 have different compute, memory, and storage profiles. Deploying AV52 nodes (1,536 GB RAM) for workloads that only need 768 GB wastes 50% of your memory allocation — and you pay for the entire node. Use the Azure Migrate assessment tool to determine which node type matches your workload profile. Over-provisioning by one node type level can cost $50,000–$100,000/year per node in unnecessary infrastructure.
Reduce node count through external storage. AVS node count is often driven by storage requirements rather than compute requirements. If your workloads are storage-heavy, attaching Azure NetApp Files or Azure Elastic SAN for data volumes allows you to run fewer AVS nodes (sized for compute) with external storage for capacity. This can reduce node count by 20–40% in storage-intensive environments, saving $200,000–$800,000/year in infrastructure costs — even after the external storage charges.
Negotiate the Broadcom VCF subscription aggressively. The portable VCF subscription for AVS is the same product as on-premises VCF licensing. All the same negotiation tactics apply: volume discounts (20–40%), multi-year commitments, scope challenge (ensuring you are licensing only the cores actually deployed, not over-provisioning for future growth). If you have both on-premises and AVS deployments, negotiate a single VCF agreement covering both environments with total core count discount tiers. See our Broadcom negotiation strategy guide for the full playbook.
Plan your exit. AVS should include an explicit exit strategy in your business case. Whether the exit is to Azure-native (re-platforming), to on-premises (if cloud costs escalate), or to another provider, the migration tools (HCX, Azure Migrate) and the contractual flexibility (RI term length, Broadcom VCF term alignment) should support the exit. Locking into a 5-year Azure RI and a 5-year Broadcom VCF subscription simultaneously creates a 5-year commitment that is difficult and expensive to unwind.
Organisations with an immovable data centre exit deadline (lease expiration, facility closure, M&A consolidation) and a large VMware estate that cannot be re-platformed in time. AVS provides a cloud landing zone that accepts VMware VMs with minimal modification, allowing the data centre exit to proceed on schedule while re-platforming continues after migration.
AVS as a DR target for on-premises VMware environments is one of the most cost-effective use cases. DR nodes are provisioned only when needed (or kept at minimum 3-node baseline), and VMware SRM/HCX replication ensures rapid failover. The cost of maintaining a 3-node AVS cluster for DR ($140K–$170K/year all-in) is significantly less than building and maintaining a secondary physical data centre.
Organisations already committed to Azure as their primary cloud platform who also have VMware workloads that cannot be re-platformed gain value from AVS as the VMware component of a broader Azure estate. Azure native services (Azure AD, Azure Monitor, Azure Backup, Microsoft Defender for Cloud) integrate with AVS, providing a unified operational model across native and VMware workloads.
If the primary motivation is reducing VMware costs, AVS is unlikely to deliver savings. You still pay for VCF licensing (same cost as on-premises), plus cloud infrastructure at a premium over self-managed hardware. AVS trades capital expenditure for operational expenditure and infrastructure management for managed service — but it does not reduce VMware licensing costs. For cost reduction, evaluate Nutanix as an alternative platform or negotiate the on-premises VMware renewal using independent advisory support.
AVS requires a minimum of 3 nodes. A 3-node AV36P cluster provides approximately 108 cores and 2.3 TB of RAM — vastly more capacity than a 50-VM environment requires. You cannot purchase fractional nodes. For small environments, the minimum cluster cost ($139,000/year all-in) produces a per-VM cost that is 3–5× higher than equivalent Azure native VMs. Direct-to-Azure migration is almost always more economical for small VMware estates.
If your workloads will run unchanged for 5+ years with predictable capacity, on-premises infrastructure with a negotiated Broadcom VCF agreement will deliver lower 5-year TCO than AVS. The cloud premium (dedicated bare-metal in Azure costs more than self-owned equivalent hardware over a 5-year period) is justified only when it is offset by data centre exit savings, staffing reduction, or agility requirements that on-premises cannot match.
Minimum ~$11,600/month for a 3-node AV36P cluster (3-year RI) plus ~$2,400/month for VCF licensing = ~$14,000/month all-in. A mid-market 10-node deployment runs approximately $38,000/month. Enterprise 30-node deployments reach $150,000+/month. These figures include Azure infrastructure (RI pricing) and Broadcom VCF subscription but exclude data egress, external storage, and ExpressRoute connectivity, which add $1,000–$10,000+/month depending on usage patterns.
Yes, as of October 2025. All new AVS node purchases require a portable VMware Cloud Foundation subscription purchased directly from Broadcom or a Broadcom channel partner. The VCF subscription is licensed per physical core and billed separately from the Azure infrastructure. Customers with Reserved Instances purchased before October 15, 2025 can continue with VCF included until their RI term expires. PAYG customers with VCF included must provide a Broadcom VCF licence key by November 1, 2026.
Rarely, for steady-state workloads. AVS eliminates capital expenditure and data centre operational costs, but the Azure infrastructure premium typically exceeds those savings over a 5-year horizon for stable environments. AVS is economically advantageous when the primary driver is data centre exit, DR, or workload variability — not VMware cost reduction. For VMware cost reduction specifically, evaluate Nutanix migration or negotiate the on-premises Broadcom renewal with independent support.
Yes, if they are portable VCF subscriptions. Broadcom’s portable VCF licence model allows the same subscription to cover on-premises hosts, AVS nodes, and other hyperscaler VMware environments. The total registered cores across all environments must not exceed your Broadcom entitlement. Legacy perpetual licences are not eligible for AVS BYOL — only current VCF subscription licences qualify. If you hold perpetual licences, you will need to purchase a VCF subscription to use AVS.
Only if direct-to-native migration is not feasible in your timeline. The “two-hop” strategy (on-prem → AVS → Azure-native) costs more than direct migration over a 5-year period because you pay full AVS rates during the intermediate period. If your technical capacity and timeline allow direct migration to Azure VMs or PaaS, that path delivers lower 5-year TCO. AVS as a stepping stone is justified when data centre exit timelines are immovable and workload complexity prevents direct re-platforming within the required window.
3 nodes per private cloud cluster. A minimum 3-node AV36P cluster provides 108 cores, 2.3 TB RAM, and approximately 57 TB raw NVMe storage. This is a substantial resource allocation that is appropriate for 100+ VMs. For smaller environments (<50 VMs), AVS is typically over-provisioned and uneconomical — Azure-native IaaS is a better fit. Up to 16 nodes per cluster and 12 clusters per private cloud are supported for larger deployments, with a maximum of 96 hosts per private cloud.
Redress Compliance provides vendor-neutral assessment of all three paths with no Microsoft, Broadcom, or Nutanix partnership and no commercial relationship with any vendor.
Explore related Broadcom/VMware licensing and advisory guides: