On November 12, 2026, Microsoft 365 users began having trouble opening their Outlook mailboxes. Some emails arrived late, and people using Teams, Copilot, SharePoint, and other Microsoft 365 products also reported problems. Outlook was the most visible part of the outage for many businesses, but the disruption started in infrastructure that several Microsoft services relied on.
Microsoft’s preliminary post-incident report points to a software defect in a shared network gateway. Teams traffic had moved onto the gateway used by Exchange Online while part of its server capacity was already unavailable. When servers began to fail, automatic rerouting sent more work to the servers still responding. That spread the failure and made recovery harder.
What caused the Microsoft 365 outage?
A gateway software defect went unnoticed for months
A network gateway handles connections between users and the services they want to access. In this case, Microsoft had started rolling out a Windows update specific to its Microsoft 365 gateway servers in July. It ran for more than three months without visible problems. Under certain loads, however, work inside a Windows process became backed up, consuming processing power and memory until servers stopped responding.
Microsoft says its load tests used a different mix of traffic and generally less powerful hardware than some of the machines running the gateway in production. The defect was more likely to appear on newer servers with more processor cores. The tests missed the combination that eventually caused the failure.
Teams traffic moved onto the gateway used by Exchange Online
About 15 hours before the outage, Microsoft’s automated failover system moved Teams traffic from an alternate gateway onto the one that also handled Exchange Online connections. Failover is a routine way to keep a service running by moving its traffic to other infrastructure. In this case, the move increased demand and changed the type of traffic arriving at the shared gateway.
Nearly 14% of that gateway’s capacity was already out of service because of an unrelated issue. The gateway had handled the capacity loss on its own. When the extra Teams traffic and the reduced capacity coincided with peak U.S. demand, the conditions exposed the software defect.
Rerouting traffic spread the failure
As servers stopped responding, Microsoft’s traffic management system redirected users to healthy servers. Each round of rerouting put more demand on a smaller pool of machines that could fail in the same way. Safeguards intended to contain failures within different parts of the gateway did not stop the outage from spreading across regions. At the worst point, more than 80% of the gateway’s servers were unresponsive, according to Microsoft’s preliminary report.
Why were Outlook and Teams affected differently?
Exchange Online, the service behind Microsoft 365 business email, took the biggest hit. The connection problems affected Outlook desktop, Outlook on the web, and other ways of reaching the same mailbox. That meant switching from the desktop app to a browser could still leave users facing the same underlying service problem. Email delivery also slowed, leaving messages waiting to be processed.
Microsoft could move Teams away from the faulty gateway. The company says Teams chat and meetings worked for most of the incident, although features that depended on Exchange Online could still be affected. Some features in Copilot, SharePoint, OneDrive, Defender, and Purview experienced similar knock-on effects. Microsoft says Exchange Online could not move away from the affected gateway at the scale required.
Why did Microsoft 365 recovery take so long?
When restarted servers came back online, waiting requests could overwhelm them again. Microsoft’s usual automated recovery processes were designed for more limited failures. Rolling back a recent deployment also offered little help because the defect had been in production for months. Engineers prepared a fix, took groups of servers out of service, applied the update, and restarted the servers before gradually returning them to use.
Restoring access and clearing the mail backlog were separate jobs. During recovery, Microsoft extended the email timeout window to 48 hours to help preserve delayed messages while it processed the queue. Some users could open Outlook again before all the messages sent during the outage reached their inboxes. BleepingComputer reported Microsoft’s update on the email backlog.
Microsoft 365 outage timeline
Date and timeWhat happened November 12, 4:20 p.m. UTCMicrosoft records the start of the Exchange Online incident under EX1550543. November 12, 6:27 p.m. UTCMicrosoft posts the broader Microsoft 365 advisory under MO1550821. November 13, 4:10 p.m. UTCThe preliminary report records the end of its incident window, just under 24 hours after it began. November 14Microsoft publicly confirms it restored service to the remaining affected customers and regions. November 15Microsoft dates its preliminary post-incident report, detailing the gateway defect and the conditions that exposed it.
The November 13 timestamp is the incident end time recorded in Microsoft’s preliminary report. The November 14 statement marks its public confirmation for the remaining affected customers and regions. Reuters covered Microsoft’s recovery announcement.
What SaaS teams can learn from the Microsoft 365 outage
Test failover with less capacity than you expect to have
A backup can handle a traffic shift in a routine test and still struggle during a real incident. Run a test with some servers already offline, then shift traffic from another service into the same environment. Watch how response times, errors, and queues change as the backup takes on more work. That tells you how much capacity you actually have when several problems overlap.
Keep the core workflow available when a dependency fails
In a booking app, the reservation and its confirmation email can run as separate steps. If the email provider is down, the app can still save the booking and show its reference number. When email service returns, a background process can send delayed confirmations at a controlled rate while new bookings continue. That separation matters in multi-role products such as the fitness marketplace in our FitFinds case study, where customers, merchants, and administrators use different parts of the same platform.
Report recovery in terms customers can act on
A status update should tell customers which actions work and where they may still see delays. For a booking platform, that might mean confirming that reservations are saving while email confirmations are still catching up. For a marketplace, it might mean that customers can place orders while some payment records are still being reconciled. Define those milestones early in SaaS development, so your team can measure and communicate them during an outage.
How to check Microsoft 365 service health
If Outlook or Teams stops working on a work account, a Microsoft 365 admin can open Health → Service health in the admin center to check for incidents affecting the organization. Microsoft’s Service Health guide explains where to find active incidents and advisories. If the admin center is unavailable, use Microsoft’s public service status page. For the November outage, the incident IDs are MO1550821 for Microsoft 365 and EX1550543 for Exchange Online.


