Microsoft Turns On Confidential Live Migration for Azure, Ending the Patch-and-Reboot Tradeoff for Secure VMs
Microsoft has switched on Confidential Live Migration for Azure, a capability that lets a protected virtual machine move from one physical host to another without exposing its memory or halting the guest for a full restart. The company showed the feature at its Build 2026 conference and began rolling it out to production for Intel TDX confidential VMs at the end of June.
For years, the strongest isolation Azure offered came with a hidden tax: applying a security patch to the host often meant rebooting the confidential machine. Microsoft says that tradeoff is now gone for eligible workloads.
What Microsoft shipped
Confidential Live Migration targets Azure confidential VMs, which run inside hardware-based trusted execution environments. On Intel hardware, that means TDX, short for Trust Domain Extensions. The TEE keeps the VM's RAM and CPU state encrypted even from the cloud provider's own hypervisor and host administrators.
Live migration is old news for ordinary VMs. Providers use it to drain a host for maintenance, balance load, or absorb hardware failures with only a brief pause. Confidential VMs could not join that dance without dropping their protections. Microsoft's new flow keeps the encryption boundary intact while the machine changes homes. Microsoft laid out the internals in its announcement on the Azure Confidential Computing blog.
Trusted execution environments flip a long-standing assumption about cloud trust. Customers already encrypt data at rest on disk and in transit across the wire, but the RAM and CPU registers of a running machine stay readable to the host. A TEE closes that gap by encrypting memory and shielding CPU state, so even a compromised hypervisor or a rogue administrator with physical access sees only ciphertext. Microsoft documents the model in its Azure confidential VM overview.
Why confidential VMs were stuck with reboots
Standard Azure VMs handle most host servicing with a short pause that lasts seconds. Confidential VMs told a different story. Until now, certain updates to the host operating system or BIOS demanded a planned reboot of the guest. Those reboots landed several times a year, each costing minutes of availability.
The math of the old model added up to real risk. A planned reboot a few times a year might sound minor, yet each one interrupted the workloads customers most wanted always on: payment systems, medical records, identity services. Worse, patching delays invite known exploits, so teams faced a choice between an outage and an exposure window.
That gap mattered most for finance, healthcare, and government, where data-in-use protection is not optional. A shop that paid for confidential computing to keep sensitive records out of the hypervisor's reach was then forced to pick between patching on time and staying online. Microsoft frames the fix as removing a maintenance downtime tradeoff, letting the platform move the VM to an updated host while preserving the isolation that justified the confidential VM in the first place.
How the migration keeps memory encrypted
The workflow starts with attestation. Azure verifies that the destination host matches a required measurement and satisfies the migration policy before any state leaves the source. Only after that check does the platform set up a protected session, perform an authenticated key exchange, and derive ephemeral keys tied to the attested destination.
Next comes the copy. Most memory pages stream across an encrypted channel while the VM keeps running. The CPU and the I/O memory management unit track which pages changed during the transfer using secure dirty-bit mechanisms, so nothing falls through the cracks. A final, brief switchover pause sends the remaining deltas, enforces a one-way handoff, and resumes the guest on the new host.
Under the hood, Intel's TDX live migration design introduces a small, open-sourced module called MigTD. MigTD sits inside the confidential trust boundary and makes the policy decisions for the move, checking the destination against the rules and folding its own measurements into the attestation evidence a remote party can verify. Microsoft says Azure's control plane coordinates the steps while MigTD and the TDX module enforce attestation, key exchange, and the one-way switchover. The MigTD code is published on GitHub for inspection.
Ordinary live migration has run underneath Azure for years, quietly shuffling standard VMs during maintenance windows. The confidential variant required new cryptography because the memory being copied is the very thing that must stay secret. Microsoft had to prove the destination was trustworthy before a single page moved, then keep the transfer encrypted end to end.
What it means for regulated and edge workloads
Confidential computing has moved from a security-lab curiosity to a line item in enterprise buying guides, pushed by tighter rules on where data may sit and who may touch it. Live migration removes one of the last operational reasons to avoid it: the fear that staying patched meant staying down.
The change also matters at the edge. As inference and sensitive processing move closer to where data is born, on factory floors, in clinics, and at branch offices, the same confidentiality guarantees that protect a VM in a regional data center now need to survive host maintenance on smaller, more fragile sites. Microsoft's approach keeps those protections portable across hosts rather than pinned to a single machine. Our earlier reporting on AI inference reaching metro data centers covers the same pull toward the edge, and the full Cloud & Edge Computing section tracks the trend.
Security teams can read more in our cybersecurity coverage, where confidential computing often shows up as a control for regulated workloads.
The road ahead
Microsoft lists attestation and policy enforcement, confidentiality, and integrity as the three design goals it will not compromise. The company points to industry collaboration as the next step, noting that confidential computing depends on silicon vendors, hypervisor builders, and cloud operators agreeing on shared attestation formats.
The first production wave targets Intel TDX VMs. The underlying TDX migration architecture and the MigTD module are open source, which lets customers and auditors check the claims for themselves. Organizations that already run confidential VMs can expect routine host servicing to stop requiring planned reboots, while new adopters get one fewer excuse to keep sensitive workloads on bare metal.
The bottom line for cloud operators is plain: the most secure VMs on Azure no longer have to choose between a patch and an outage.