The guide is useful, though it blurs some details. Microsoft has since published its own post-incident reports, and they correct and add to the popular account. Here's what happened, what Microsoft says caused it, and how Microsoft 365 shops can prepare for the next one.
What actually broke on November 12–13
The headline window is now settled. According to Microsoft's preliminary Post Incident Report, as summarized by IncidentHub, Exchange Online, and parts of the Microsoft 365 services that depend on it, were disrupted worldwide from 16:20 UTC on November 12 to 16:10 UTC on November 13, 2026 - 23 hours 50 minutes. BleepingComputer reported that in US time the incident began at 11:20 AM ET on Thursday, November 12, and continued until 11:10 AM ET the following day.
Exchange took the worst of it. The preliminary report, reproduced by Office Watch, says the most significant impact affected Exchange Online, with most users being unable to access their mailbox through Outlook or other mail clients. Additionally, email delivery was delayed for many users. The admin-center tracking IDs are EX1550543 and MO1550821.
The damage spread well beyond Outlook. 365Fanatic summarizes Microsoft's final report, which lists Exchange, Copilot, Teams, SharePoint, OneDrive, Defender, Purview, Fabric, Intune, Planner, Power Apps, To Do, Power BI and Universal Print. In that summary, Teams, SharePoint and OneDrive recovered earlier because their traffic was moved off the failing gateway. Features that rely on Exchange data, such as search, calendar and presence, stayed broken for longer.
On scale, IBTimes notes that Microsoft 365 has about 430 million users and that problems were reported in more than 70 countries. The 430 million figure is the total user base. It is not a confirmed count of affected users.
IBTimes also cites three claims that have not been independently confirmed:
- an Anderson Economic Group loss estimate of about $7 billion
- cancelled operations and disrupted urgent referrals in British healthcare
- UK minister Luke Pollard's call for "greater resilience"
Read these as IBTimes' reporting.
Section summary: The outage lasted nearly 24 hours and centered on Exchange Online. It also hit Copilot, Teams, security and compliance tools, Intune and Power Platform services.
The cause: four ordinary conditions at once
IBTimes describes three factors: a hidden software defect, extra traffic from a routine network change, and a cascade in which isolation barriers failed. That roughly matches the early version. In its preliminary report, BleepingComputer wrote, Microsoft named three problems: a previously undetected software bug, increased traffic after a Microsoft Teams failover, and reduced server capacity.
The final report, released on November 20 and summarized by 365Fanatic, lists four factors. Each is harmless alone:
- A latent defect. An update to the Microsoft 365 networking gateway servers had been rolling out since July. Under certain load, tasks inside a Windows process waited on each other, which exhausted CPU and memory until servers stopped responding. The defect hit only newer, high-core-count hardware. Microsoft's test labs used older machines with different traffic patterns, so normal load testing missed it.
- Reduced capacity. Nearly 14% of gateway capacity was out of service as part of routine deployment work.
- A Teams failover. A routing problem in US East moved Teams traffic onto the gateway that also serves Exchange Online.
- Peak load. Broad failures began around 16:00 UTC, as US traffic approached its daily peak.
Then the cascade started. Traffic management did what it was designed to do and moved load from failed capacity units to healthy ones. The extra load pushed those units into the same defect. Microsoft's report says more than 80% of North American gateway servers were unresponsive at the worst point, and the problem reached every region.
Recovery was slow for a reason that should worry every change-management team. The defect came from a change rolled out months earlier, so the usual fixes (halting the rollout, rolling back, or turning off a feature flag) couldn't undo it quickly. Each server had to be taken out of rotation, restarted, patched, restarted again and returned to service.
A note for Windows admins: the bug sat in a Windows process. According to the final report as summarized by 365Fanatic, neither the defect nor the update applies to broader Windows deployments. Your own servers are not affected.
On whether this was an attack: Microsoft said during the incident that it was "not aware of any malicious activity causing this incident". 365Fanatic's summary of the final report adds that the defect-causing change was not made by AI tooling.
Section summary: No attacker was involved. A months-old latent bug, reduced capacity, a traffic shift and peak demand combined, and automatic failover spread the failure globally.
The communication trap: your alerts may go to the broken service
This is the most useful practical point, and IBTimes doesn't cover it.
Microsoft's documentation lets admins subscribe to Service health email notifications for up to two addresses per admin. Many organizations point those alerts at their own Exchange Online mailboxes. In this incident, the "Exchange is down" email would have gone to Exchange.
Microsoft lists several other status channels. Its Learn documentation says:
- If you can't sign in to the admin center, check the public service status page.
- Follow @MSFT365Status on X.
- Install the Microsoft 365 Admin app for push notifications.
Microsoft's service description also lists the Service Communications API in Microsoft Graph and a System Center Operations Manager management pack for teams that want status feeds in their own tooling.
Microsoft 365 status-channel checklist:
- Set at least one Service health notification address that is not hosted on your Microsoft 365 tenant.
- Make sure several people can read Service health. Microsoft notes that the Service Support admin and Helpdesk admin roles can view it.
- Install the Admin app on on-call phones before you need it.
- Write down the public status page and the X account in your printed runbook.
- Learn Microsoft's labels. An advisory means some users are affected but the service is still available. An incident means the service or a major function is unavailable. Statuses run from Investigating to Service restored, with Extended recovery in between.
Section summary: Monitoring doesn't prevent downtime. Alert channels that run through the failed service are a weakness, so store your status sources outside the system you're monitoring.
IBTimes' six steps, with Microsoft 365 detail
1. Map dependencies. IBTimes says to list systems you can't operate without, such as email, payroll, point of sale, customer databases and phones, and note who runs each one. This outage showed why hidden links matter. Intune, Purview, Defender and Universal Print all lost functions during an outage that was mostly about Exchange. If printing, device management or your security operations center depends on Microsoft 365, record that before it breaks.
2. Build backups and alternatives. IBTimes recommends keeping critical data separate from the main provider and having fallbacks such as a backup email system, a separate messaging tool, or a phone tree that works without the internet. Check that a fallback is truly independent. A Teams backup for Exchange email would have failed early in this incident, because both share the same gateway and sign-in. Spreading workloads across clouds or regions adds cost and complexity, as IBTimes acknowledges. Multi-region designs within one provider also have limits: here, the defect was already on every region's servers.
3. Plan to work offline. Printed contact lists, paper forms and pre-approved manual processes only help if staff know where they are and have practiced using them. A binder nobody has opened won't save you.
4. Write a communication plan. Decide ahead of time how you'll reach staff and customers: SMS, a status page hosted outside your main provider, or social media. Name a spokesperson and set an update schedule. Microsoft's service description says it updates active incidents hourly unless it states otherwise, which is a reasonable model for your own updates.
5. Review contracts and insurance. IBTimes correctly notes that SLA service credits rarely cover real business losses, and that business-interruption and cyber policies vary widely. Microsoft's service description says the SLA continues to apply while redundancy is being rebuilt after a catastrophic outage. Check what your own agreement and policy actually say. Neither this article nor IBTimes is legal or insurance advice.
6. Test and update. IBTimes says drills should simulate a major outage, followed by fixes. Microsoft's report makes the same point about its own testing: its lab conditions didn't match production, and the bug survived. A tabletop exercise based on "Exchange is down for 24 hours, starting at 11 a.m. Eastern" would be a good start.
After the incident: what Microsoft owes you
Microsoft's service description says broad incidents get a preliminary post-incident review in Service health within 48 hours of resolution and a final review within five business days. Both timelines were met here: the preliminary report came on November 15 and the final on November 20. Azure Blog notes the report can also be downloaded with Graph: GET /admin/serviceAnnouncement/issues/EX1550543/incidentReport.
Two other mail-flow details are worth knowing:
- According to the final report as summarized by 365Fanatic, no mail was lost. Microsoft extended Exchange Online's retry window from 24 to 48 hours.
- Delayed mail to consumer mailboxes that had been inactive for more than 30 days did expire, and the senders received non-delivery reports. If customers say their message bounced during the outage, that may be why.
Microsoft's planned fixes include:
- Splitting gateway capacity into pools that serve different workloads (target: Q1 2027)
- Limiting Teams failover to the affected region instead of moving it worldwide (target: Q4 2026)
- Load testing on the newest high-core hardware with production-like traffic (target: Q1 2027)
- Adding capacity headroom (target: Q1 2027)
Some repair items still have no date.
The bottom line
IBTimes is right that no business can prevent a cloud outage, and that planning can turn a shutdown into a delay. The Microsoft 365 incident also shows how such failures happen: through normal events lining up badly, not through one dramatic error. A bug that ran safely for more than three months failed only when reduced capacity, a failover and the US morning peak hit at the same time.
What you control is your own preparation: an honest dependency map, a fallback that doesn't share the same failure point, a printed runbook, and status alerts that reach someone even when Exchange is down. Has your team checked where its own alerts go?
References
- When the Cloud Goes Down: How Businesses Can Prepare for the Next Major IT Outage Before It Hits Them - International Business Times Australia International Business Times Australia · 2026-11-22T04:28:23+00:00
- Microsoft 365 outage 12 November 2026: final report | 365Fanatic 365fanatic.com
- Microsoft 365 outage 12 November 2026: post-incident report 365fanatic.com