NEWS
Microsoft 365 Outage Spreads From Exchange to Copilot
A health-check failure on Microsoft 365 servers on November 12 knocked Outlook offline and cascaded into Copilot, Defender, Intune and third-party AI connectors.
Microsoft 365 servers began failing automated health checks on Thursday, November 12, and Downdetector logged a peak of 138,058 Outlook reports around 1 p.m. Eastern. The company later said traffic had not risen. A process that sets up secure connections was overwhelming the remaining machines, and Copilot, Defender, Intune and third-party AI connectors went with the mail.
Microsoft was still posting service degradation early on November 13. A spokesperson said the company had no indication of a cyber-attack.
Outlook Peaked at 138,058 Reports
Downdetector, which compiles user outage reports, said the trouble reached 74 countries in an hour. Outlook reports ran more than 2,000 times their usual Thursday level. Between 17:00 and 18:55 UTC, nearly 3 in 10 of those reports came from outside the United States.
Microsoft opened Exchange Online incident EX1550543 at 11:17 a.m. Eastern (16:17 UTC) and a broader Microsoft 365 incident, MO1550821, at 1:26 p.m. Eastern. The first public note on X went out at 12:54 p.m. Eastern and named mailbox access in the Americas. Thirteen minutes later the company said users in any region could be hit.
THE DOWNTIME PEAK
- Outlook reports: 138,058 around 1 p.m. Eastern on November 12, falling to 34,490 by evening, per Downdetector.
- Five-service burst: Combined reports on five Microsoft products hit 19,650 a minute, against fewer than 15 on an average Thursday.
- By 5:47 p.m. Eastern: Microsoft 365 reports were nearly 13,000 and Teams reports were 763.
- Azure contrast: Azure reports rose about a hundredfold, while Xbox, GitHub and LinkedIn rose far less.
The North Dakota University System and a Japanese research-university help desk both relayed Microsoft’s incident notes to staff, a sign of how widely education tenants lean on the same pool.
What Microsoft Has Confirmed About the Cause
Microsoft 365 Status, the company’s incident account, started with mailbox access and kept widening the picture as the afternoon ran on.
We're investigating reports of users experiencing issues accessing Exchange Online mailboxes through multiple connection methods in the Americas. More details can be found in the Microsoft 365 admin center under EX1550543.
— Microsoft 365 Status (@MSFT365Status) November 12, 2026
The diagnosis moved in public. Elevated load came first. Then Microsoft said the load was not from more traffic. Servers had dropped out of service, and the remaining pool was taking the strain. After that came failed automated health checks, a core process that would not keep up, and a secure-connection process that stopped machines from handling requests.
HOW MICROSOFT’S STORY CHANGED
- November 12, 9:45 a.m. ET: Separate Teams incident TM1550488 hits North and South America, with message delays and chat errors, and is marked resolved at 10:53 a.m. ET after 68 minutes. Microsoft blames resource contention and does not tie it to the later outage.
- November 12, 12:54 p.m. ET: Investigates Exchange Online mailbox access in the Americas under EX1550543.
- November 12, 1:07 p.m. ET: Says the issue may affect Microsoft 365 access for users in any region.
- November 12, 1:35 p.m. ET: Points to elevated load across multiple regions and opens MO1550821 beside EX1550543.
- November 12, 2:31 p.m. ET: Says the load is not from more traffic but from servers that became unexpectedly unavailable.
- November 12, 2:50 p.m. ET: Tests a configuration change for servers that fail automated health checks, and restarts the affected machines.
- November 12, 3:19 p.m. ET: Says it has not identified a recent change correlated with the start of impact, and that failed health checks may come from a component that processes server health.
- November 12, 5:11 p.m. ET: Says actions so far have not provided the anticipated relief, and that a core process on affected servers is becoming overwhelmed.
- November 12, 8:06 p.m. ET: Names a process that establishes secure connections, begins reverting a recent network configuration change that correlates with the start of impact, and limits some traffic types.
- November 13, 3:13 a.m. UTC: Says changes applied to improve serviceability are not taking as expected, and that it is identifying other ways to add capacity.
Microsoft paused all deployments and configuration changes while it looked for a correlation. It pointed customers to the public Microsoft 365 status page if the admin center would not load, which for a stretch it did not.
Servers Failed Their Own Health Checks
The mechanical picture that emerged is a death spiral, not a traffic jam. Machines failed automated health checks and left the pool. Load shifted onto what remained. A core process, later described as the work of setting up secure connections, became overwhelmed, and more servers stopped responding.
Microsoft tried an alternative health-check method on a subset of machines and said that subset came back. It prepared an automated recovery job to detect and restart servers that went quiet. It also began rebuilding some machines on a previous software configuration. None of that held once live traffic returned.
The servers we returned to service can become unresponsive again once they receive traffic; we’re analyzing diagnostic data to determine what’s triggering the behavior.
Microsoft 365 Status, on X, November 13, 2026
In a later note the same account said the servers can become unresponsive again under traffic, and that it was limiting which request types could reach the damaged pool in some locations. By 3:13 a.m. UTC on November 13 it said the serviceability changes were not taking as expected.
Every Tool That Still Talks to Exchange
Overnight on November 12, Microsoft narrowed the blast radius in a useful way. Features that interact with Exchange Online and email were hit. Parts of those products that do not talk to Exchange still worked. That is why a Teams call could go through while a calendar invite would not, and why a file sitting in SharePoint could open while Copilot could not read the inbox beside it.
Microsoft’s own resiliency design notes say an Exchange issue should not hit core Teams functions, and calls, meetings, group chats and channel messages did keep working. Calendar, scheduling, live events, Copilot work data, Defender Safe Links, Intune enrolment and several admin portals did not. The isolation held where Microsoft said it would. The longer list is everything that still waits on mail.
| Service | What broke | What still worked |
|---|---|---|
| Exchange Online | Mailbox access in Outlook on the web, desktop MAPI, ActiveSync and REST; delayed delivery | Local copies in classic Outlook Cached Exchange Mode |
| Microsoft Teams | Calendar, meeting scheduling, live and virtual events; Information Barriers chats | Calls, meetings, group chats and channel messages |
| SharePoint Online | Features that depend on Exchange | Files stored directly in SharePoint |
| Microsoft 365 Copilot | Copilot Chat in web and desktop; reach into email, calendar and files | Microsoft did not list a working Copilot path |
| Defender for Office 365 | Safe Links may be unable to verify URLs | Microsoft did not list a working Safe Links path |
| Microsoft Intune | Device enrolment, check-in and compliance policies | Microsoft did not list a working Intune path |
Microsoft later added Purview portal pages, Planner, To Do, Power BI sensitivity-label exports, Universal Print, Microsoft 365 Apps admin center, Education Assignments, Graph requests for mail, calendar and contacts, and Windows 365 Cloud PCs set up with Microsoft Purview Customer Key. Some admin center pages would not load, which left tenant admins reading X and the public status board instead of their own health dashboard.
OpenAI and Anthropic Lost Their 365 Connectors
The same Exchange dependency showed up outside Microsoft’s own apps. Workplace AI tools that read mail and calendars through Microsoft 365 connectors failed in step with Outlook, which is the second-order hit for any company that put Copilot-class assistants on top of the same mailbox.
WHO ELSE LOST THE LINK
- OpenAI: The company said it was looking into Microsoft connectors, with new connections possibly broken and tools not working as expected.
- Anthropic: Claude’s status board, from 17:52 UTC on November 12, said an upstream provider issue was hitting the Microsoft 365 connector, and that mail, calendar and search tools were failing for most requests, including Claude for Outlook.
- Microsoft Graph: Requests for mail, calendar and contacts were listed among the Exchange-tied failures, which is the pipe those connectors use.
- Customer inboxes: Firms whose support queues live in Exchange had no outbound mail, even when internal Teams chat still worked.
That split is the practical lesson. An office can keep talking to itself on Teams and still be closed to customers, vendors and ticket systems until Exchange comes back. Small support desks found that out in public on November 12.
The Last Major 365 Outage Ran Into September
This was not the first time a shared Microsoft 365 component took a long list of products with it. Microsoft’s service health record for the late-August incident lists a start of 7:56 a.m. PDT on August 31, 2026, and an end of 3:00 a.m. PDT on September 3. That event was tracked as EX1464935, then as MO1465074, and Microsoft blamed a core authentication configuration used by multiple Microsoft 365 services.
The September failure also spread from Exchange into Teams, SharePoint, Copilot, Purview, Defender XDR, OneDrive for Business, Universal Print and the admin center. Mail flow came back first. Search lagged. The shape is familiar: one shared layer, a suite-wide ticket, and a recovery that does not land in a single restart.
On November 12 the shared layer was capacity and health, not authentication. Microsoft has not published a root cause for the newer incident, and it has not said when service will be whole. Exchange Online’s published SLA still sets a 99.9 percent monthly uptime target, with service credits starting at 25 percent of the applicable fee if that mark is missed. Credits are not damages, and they do not cover lost revenue. Admins who want the official incident trail can follow Microsoft’s guide on how to check Microsoft 365 service health once the admin center itself will load.
Why Restored Servers Died Under Traffic
The last solid technical note is the one that explains why the afternoon’s restarts did not close the ticket. Servers came back clean, then went quiet again when they began receiving traffic. Microsoft isolated a slice of the service, restarted it, and reintroduced load in steps to see if the slice would stay up. It limited request types in some locations. It rolled back a network setting that lined up with the start of impact. Early on November 13 it said those serviceability changes were not taking as expected.
Until that loop breaks, bringing a machine back is not the same as keeping it. Clients will keep retrying, and those retries are themselves load on a pool that already could not finish a secure handshake. Microsoft told customers that as capacity returns, recovery may look gradual as apps reconnect. It has not named the component inside the health-check path, and it has not given a time for a full restore.
The November 12 outage started as mailbox errors in the Americas and ended the UTC day as a still-open, multi-product degradation whose working theory is a health system that removed the capacity it was meant to protect. Copilot, Defender and Intune were not random extras on the status board. They were the rest of the workday that still has to ask Exchange for permission to proceed.
-
BUSINESS3 months agoConsumer Sentiment Falls to 51.7 as Future Outlook Darkens
-
GAMING3 months agoTake-Two Subpoenas Microsoft While GTA 6 Preorders Soar
-
NEWS3 months agoPluto’s Heart Glacier Still Pushes Liquid Nitrogen Upward
-
NEWS3 months agoThe Extra Heat Month Hits Children Where Care Is Thinnest
-
ENTERTAINMENT3 months agoNetflix Weighs Hosting Peacock and Fox One in Its App
-
NEWS3 months agoFDA Clears Rasonque and Widens Who Can Take It
-
NEWS3 months agoUMMC Will Rebuild 30-Year-Old Cancer Labs With $2.4 Million
-
GAMING3 months agoGTA 6’s 80-Hour Run Counts Goals That Change the Story
