On August 18, Microsoft quietly announced that the license-included tier of Azure VMware Solution (AVS) will retire on August 30, 2027. Sales of new license-included nodes stop earlier — October 31, 2026. The Register picked it up two days later, correctly framing it as what it is: Broadcom's VCF-or-nothing licensing finally closing the last practical side door on Azure.
It reads like a routine SKU retirement. It isn't. For every organization that adopted AVS specifically to keep running VMware without buying into Broadcom's Cloud Foundation bundle, this is the moment the decision they've been deferring gets a calendar date.
What was actually announced
The mechanics, per Microsoft's update and confirmed by The Register:
- October 31, 2026: Microsoft stops selling license-included AVS nodes — PayGo or Reserved Instance.
- August 30, 2027: license-included AVS is fully retired. Existing nodes stop working under the current licensing model.
- The exception: customers who bought Reserved Instances on or before October 15, 2025 can keep running unchanged until the 2027 deadline.
- The replacement path: AVS VCF BYOL — bring your own Broadcom-issued VCF license (vSphere, vCenter, vSAN, NSX) and run it inside Azure's infrastructure.
Microsoft's own guidance is blunt about the runway: "start assessing current AVS environments immediately, and develop a transition roadmap to ensure you have ample time to exit AVS and avoid service disruption."
Why this is the last door, not just another one
Since taking over VMware, Broadcom has stopped selling a standalone, low-end virtualization SKU. vSphere Foundation on its own is barely sold at all outside of a bundle that includes the full VCF suite. AVS's license-included tier was one of the few remaining ways to consume VMware capability without personally negotiating a VCF contract with Broadcom — the hyperscaler carried the license, you just paid for consumption.
That arrangement is what's ending. Every hyperscaler VMware service has now moved, or is moving, to the same BYOL model. The Register notes this isn't Broadcom's first forcing move either — smaller cloud providers have twice been pushed to drop VMware support entirely rather than restructure around VCF.
So the honest framing for AVS customers isn't "renew or don't." It's: buy into full VCF from Broadcom on Broadcom's terms, or start planning an exit from VMware altogether. There isn't a lightweight middle path left.
Where this lands for sovereignty-focused shops
This is where the timing gets interesting for the audience I write for. A meaningful share of AVS estates in Europe sit inside organizations already re-evaluating their infrastructure for NIS2, DORA, or national baselines like the Dutch BIO — reviews that were happening anyway, on a slower clock. AVS's retirement notice hands those same teams an external, board-visible deadline to fold the VMware question into that review instead of treating it separately.
For anyone in that position, "exit AVS" doesn't have to mean "buy VCF and stay." Azure Local is the credible internal alternative: Arc-enabled VMs and AKS for the workloads that don't need vSphere-specific tooling, a control plane you can run cloud-connected or, per Microsoft's own disconnected-operations roadmap, fully disconnected for the workloads that actually require it. It's also — not incidentally — the platform Microsoft is positioning as the sovereign landing spot for exactly this kind of forced migration.
None of that makes the migration free. Moving off vSphere-native tooling has real engineering cost, and "we have to leave VMware because Broadcom changed the licensing" is not the same conversation as "we chose to leave VMware." But a forced deadline is still leverage: it's the business case that gets a VMware-exit project funded when "sovereignty" alone wasn't enough to move the budget.
What's not answered yet
Being straight about the gaps, because that's the point of this site:
- No public cost comparison. Neither Microsoft nor Broadcom has published what a standalone VCF license actually costs an AVS customer versus what license-included AVS cost. Until that number exists, "just do BYOL" isn't a real option to weigh against migration — it's a placeholder.
- No migration tooling story yet. Microsoft has Azure Migrate for AVS-to-Azure-native scenarios, but nothing specific has been published for AVS-to-Azure-Local. That gap is worth watching — and worth testing hands-on once there's something to test.
- Workload triage guidance is missing. Which AVS workloads are reasonable BYOL-and-stay candidates versus which should move is currently a judgment call, not a documented decision framework.
Who should care
Architects and infra leads with AVS in production, anyone running a NIS2/DORA/BIO-driven infrastructure review this year, and consultants who get the "should we finally leave VMware" question. The deadline is 2027, but the sales cutoff — October 2026 — is close enough that the assessment needs to start now, not next quarter.
Broadcom closed the low-cost side door out of full VCF lock-in. Microsoft's answer, deliberately or not, is the same platform it's already selling as the sovereign one.
I'm scoping a lab to actually test the AVS-to-Azure-Local path once there's enough published guidance to build on. If you want that when it lands — or you're weighing this decision right now and want a second opinion — subscribe to Sovereign Cloud Watch or find me on LinkedIn.