Microsoft has published a preliminary post-incident report for the Microsoft 365 outage that began on 12 November 2026. It traces the fault to a change that started deploying in July, which left what Microsoft calls “a latent defect” in Windows servers handling Microsoft 365 network traffic. At its lowest, Microsoft says more than 80% of those servers were unresponsive.
The question most IT managers actually have is whether their own Windows estate carries the same defect. Microsoft answers it directly: “neither the underlying defect nor the fix applies to broader Windows deployments.” This was specific to the gateway servers inside Microsoft’s own infrastructure.
“Our normal load testing did not detect the latent issue.” Microsoft, preliminary post-incident report for EX1550543, 15 November 2026.
What Microsoft's Report Actually Says
Most coverage of this Microsoft 365 outage has been written from news reports. The report itself is published to the Service Health Dashboard under incident EX1550543 and is dated 15 November 2026, and it is more specific than the summaries.
Four things lined up.
- A change from July. Microsoft says “the initial change in behavior first started deployment in July of this year.” It ran for over three months without visible problems.
- A defect nobody had found. “A latent defect caused Windows servers to become overwhelmed and unresponsive under certain load conditions,” caused by “lock contention inside an internal Windows process.”
- Reduced capacity. Microsoft reports that “nearly 14% of server capacity for the networking gateway was out of service” when the incident began.
- A surge in load. Microsoft says “a surge in load increased the likelihood for the networking gateways to encounter the contention,” and the result was “a cascading failure from server to server.”
Note what that sequence describes. No single failure, no attack, and no bad deployment on the day. A dormant defect plus reduced headroom plus a traffic spike, three months after the change that introduced it. Microsoft has been explicit that this was a technical fault rather than a security incident.
Is Your Own Windows Estate Affected?
No, and this is the part worth telling your board. When a vendor says a Windows bug took down a global service, the reasonable next question is whether the same bug is sitting in your own servers.
Microsoft addresses it in the report: “neither the underlying defect nor the fix applies to broader Windows deployments.” The defect was in the networking gateway tier inside Microsoft’s infrastructure, not in Windows Server as you run it. There is no patch for you to chase and no action for your own estate arising from the root cause.
That matters because the wrong lesson here is a patching panic. The right lesson is about dependency, not vulnerability. Nothing in your environment failed during this Microsoft 365 outage. Your access to a service you do not operate failed, and that is a different category of risk requiring a different kind of answer.
Reviewing your dependency on one provider?
Talk to the Transputec team about the resilience work behind our Managed IT Services and 24/7 UK service desk.
Get in TouchYour Security Tooling Went Down With Your Email
This is the part almost nobody has written about, and it is the most important consequence of the Microsoft 365 outage for anyone thinking about resilience.
Microsoft’s own advisory MO1550821 listed what was affected: Exchange Online, Microsoft Teams, SharePoint Online, Microsoft 365 Copilot and Microsoft Defender XDR.
Read that last pair again. The AI assistant people have spent a year embedding into daily work went down. So did the extended detection and response platform some organisations rely on to see attacks. Both failed at the same time as email, because both are tenants in the same building.
| What went down | Why it matters beyond the inconvenience |
|---|---|
| Exchange Online, Teams, SharePoint | The obvious half. Communication and documents stop. |
| Microsoft 365 Copilot | If AI is now in a workflow, that workflow has a single-provider dependency nobody priced for |
| Microsoft Defender XDR | Detection and response is a control, and controls that share a failure domain with the thing they protect are not independent |
Nothing about this suggests abandoning any of those products. It suggests knowing which of your controls would be unavailable at the same moment as your email, and deciding in advance whether that is acceptable. For most organisations it is the first time that question has had a concrete answer.
The Line Worth Reading Twice
“Our normal load testing did not detect the latent issue.”
That is Microsoft, in its own report, about its own testing. It is a more useful sentence than any vendor marketing claim, because it settles an argument that comes up in every resilience conversation.
The argument goes: hyperscale providers have more engineering capability than we ever will, so our effort is better spent elsewhere. That is largely true, and it is why most organisations are more secure and more available on Microsoft 365 than they were on a server in a cupboard. This post is not an argument against the cloud.
But the conclusion people draw from it, that provider failure can therefore be left out of the plan, does not follow. A defect survived three months of production and Microsoft’s own load testing. If that can happen to Microsoft, planning that assumes it cannot happen to your provider is planning with a gap in it. The question the Microsoft 365 outage actually poses is not whether your provider is competent. It is what you do during the hours when a competent provider is still fixing something.
What This Means If You Are FCA Regulated
Here is the angle missing from everything else written about this outage, and for regulated firms it is the one with a deadline attached.
The FCA published PS26/2, Operational incident and third party reporting, on 18 March 2026. The rules come into force on 18 March 2027. They require firms to submit standard or enhanced incident reports through a single route, and to maintain a register of material third party arrangements that is submitted to the FCA annually.
In setting those rules the FCA states plainly that many of the incidents reported to it originate at third parties. A day-long outage of the platform carrying your email, your documents and in some cases your security monitoring is exactly the scenario the regime was written for.
Which makes this week a free rehearsal. Four questions are worth answering while it is fresh.
- Would this have met your reporting threshold? If you cannot answer that, the threshold is not written down clearly enough.
- Could you produce the timeline? Detection, escalation, business impact, resolution, each with a timestamp.
- Is Microsoft on your material third party register? For most firms it plainly should be, and the register becomes an annual submission in March 2027.
- Which important business services were affected, and did you stay inside your impact tolerances? That framing is the FCA’s, not ours.
Our guide to IT support for financial services covers the wider regime. The point here is narrower: an incident you did not cause is still an incident you may have to report.
What to Actually Change
Four changes, and deliberately none of them is “buy a second email system”. That is usually the wrong answer for the cost.
- Get your alerting outside the thing that breaks. If your outage notifications arrive by email on the platform that is down, you find out from a member of staff. A phone number or an address on another domain fixes this for nothing.
- Write down the one page people need. Who decides, how you reach clients and staff without email, and where the key phone numbers live. It needs to exist somewhere that is not SharePoint.
- Move emailed sign-in codes off email. Anything that authenticates by sending a code to a mailbox fails twice over during a mail outage. Authenticator apps or passkeys remove that dependency and are better practice anyway.
- Know which controls share a failure domain. If your detection platform and your email are the same tenant, write that down as an accepted risk or do something about it. Either is fine. Not knowing is not.
None of those four would have shortened the Microsoft 365 outage by a minute. All four would have changed how your organisation experienced it. If you want the fuller treatment, our guide to creating an incident response plan and our piece on the benefits of an IT business continuity plan cover the planning side properly rather than in four bullets.
What Not to Do During an Outage
Half of the damage from a provider outage is self-inflicted, and all of it is avoidable.
- Do not reset passwords. Sign-in failures during a service fault are not credential problems. Mass resets create a second incident and a support queue that outlasts the outage.
- Do not reinstall or reconfigure clients. Rebuilding Outlook profiles while the service is down produces broken profiles once it returns.
- Do not make permanent changes under pressure. DNS edits, mail routing changes and tenant settings altered during an outage are the things still causing problems a fortnight later.
- Do not trust unsolicited help. Outages reliably attract phishing and fake support calls, because attackers know who is expecting contact from IT. Treat anything inbound as suspect.
- Do not tell staff to use personal email. It feels pragmatic and it puts company data somewhere you cannot reach, retain or protect.
During the Microsoft 365 outage the single most useful instruction to send was a short one: the service is down, we are tracking it, do not change your password or your settings, and here is where updates will come from.
Where to Check Next Time
Three places, in order, and the first is the only one that is authoritative for your tenant.
- Service health in the Microsoft 365 admin centre. Tenant-specific, and the only place the post-incident report for this event appears. It is filed under EX1550543.
- Microsoft’s public status page. Useful when you cannot sign in to the admin centre, which during a sign-in outage is exactly when you need it.
- Independent reporting. Often faster than either in the first thirty minutes, and wrong about the cause more often than not. Useful for confirming it is not just you.
One practical note on the report itself. Service health history is retained for a limited window, so if you want the post-incident report for your own records, take a copy now rather than in February. Microsoft states that “a final PIR will be provided within five (5) business days from full event resolution”, so a fuller version with the complete list of repairs is expected shortly after this was written.
How Transputec Approaches Provider Failure
Honestly: nobody keeps Exchange Online running when Exchange Online is down. Any provider claiming otherwise is describing a different product. What a managed service can do is remove the two things that turn an outage into a bad week.
The first is finding out late. Transputec runs a 24/7 UK service desk as part of Managed IT Services, which means monitoring and notification do not depend on somebody noticing their email has stopped. The second is the self-inflicted damage above, which is almost entirely a function of nobody authoritative being available to say “change nothing”.
For regulated firms there is a third: the evidence. A timeline with timestamps, an impact assessment against your important business services, and a third party register that is current rather than reconstructed. That is work that has to be done before the incident, and from 18 March 2027 it stops being good practice and becomes a reporting obligation.
If this Microsoft 365 outage cost you a day and you are not sure what you would have had to show a regulator afterwards, that is the conversation worth having. Talk to our team about the resilience and reporting work behind our Managed IT Services.
Frequently Asked Questions
What caused the Microsoft 365 outage?
Microsoft’s preliminary post-incident report, filed under EX1550543 and dated 15 November 2026, traces it to a change that began deploying in July 2026. That change left “a latent defect” which “caused Windows servers to become overwhelmed and unresponsive under certain load conditions”, through “lock contention inside an internal Windows process”. Two conditions then triggered it: nearly 14% of the networking gateway capacity was already out of service, and a surge in load made the contention more likely. The result was “a cascading failure from server to server”, with more than 80% of those servers unresponsive at the low point.
Does the Windows bug affect my own servers?
No. Microsoft states that “neither the underlying defect nor the fix applies to broader Windows deployments”. The defect sat in the networking gateway tier inside Microsoft’s own infrastructure, not in Windows Server as customers run it. There is no patch to chase and no remediation for your own estate arising from the root cause. The lesson from this incident is about provider dependency, not about a vulnerability in your environment.
Was the outage a cyber attack?
No. Microsoft has been explicit that this was a technical fault and not related to a cyber attack. That is worth communicating internally, because an unexplained day-long outage generates exactly the speculation that leads to unnecessary password resets and panicked changes. Separately, outages do attract opportunistic phishing and fake support calls, so inbound contact during an incident still deserves suspicion even when the incident itself is not an attack.
Which Microsoft 365 services were affected?
Microsoft’s advisory MO1550821 named Exchange Online, Microsoft Teams, SharePoint Online, Microsoft 365 Copilot and Microsoft Defender XDR. The last two are the ones worth noting: an AI assistant now embedded in daily work, and an extended detection and response platform some organisations rely on to see attacks, both unavailable at the same time as email because they share the same failure domain.
Do we have to report a provider outage to the FCA?
Potentially, and from 18 March 2027 the rules are firmer. The FCA’s PS26/2 on operational incident and third party reporting, published 18 March 2026, comes into force on 18 March 2027 and requires firms to report qualifying operational incidents through a single route and to maintain a register of material third party arrangements submitted annually. The FCA notes that many incidents reported to it originate at third parties, so an outage you did not cause can still be an incident you have to report. Whether this specific event crossed your threshold depends on the thresholds you have defined, which is itself the thing worth checking now.
This article was drafted with AI assistance and reviewed by the Transputec team. Featured image: AI-generated.



