Connect with us

NEWS

Microsoft 365 Is Back After a Capacity Upgrade Failed

Microsoft restored remaining 365 customers on Saturday after a capacity upgrade failed, then a slow recovery stretched the outage into a second workday.

Published

on

Microsoft said on Saturday it had restored remaining Microsoft 365 customers after a global outage that a network capacity upgrade helped to cause. Outlook, Teams and Copilot had been down since Thursday for governments, hospitals, courts and companies worldwide.

The company closed incident MO1550821 and said the fault was not a cyberattack and not related to AI agents. A problem introduced during an earlier network capacity upgrade surfaced after a later network change, then recovery traffic itself threatened to knock restored servers over again.

The Capacity Upgrade That Knocked 365 Offline

Microsoft’s first public account pointed to elevated load across multiple regions. It later said the load was not extra user traffic. A large number of servers had become unavailable, and the rest carried the strain.

The next version was more specific. A process on the affected servers that sets up secure connections became overwhelmed, the company said, and those machines failed automated health checks and dropped out of rotation. Servers brought back froze again once live traffic reached them.

Engineers paused deployments, rolled back a recent network configuration change that lined up with the start of impact, and patched an upstream authentication and routing component region by region. Microsoft said it restored service in stages, checking stability at each step, because demand was too large to bring everything back at once.

Our investigation points to technical challenges and failure modes (cascading failures + thundering herds at our network perimeter) that made recovery more difficult than normal.

Charles Lamanna, executive vice president of Copilot, Agents and Platform, on X

That is the twist inside the outage. The work meant to add network capacity introduced the fault. A later network change brought it to the surface. When machines came back, waiting clients hit them together, which is why a faster restore risked another collapse.

Steven Murdoch, a professor of security engineering at University College London, said an outage of this size is usually the result of some surprising interaction between systems that are each well understood on their own. Microsoft has promised a full post-incident report once its review is complete. Until that paper lands, the public account remains an upgrade, a later change, and a recovery that had to be throttled.

Microsoft 365 Status posted the all-clear for remaining customers on Saturday morning.

Outage Reports Hit 19,650 a Minute

Downdetector, an Ookla service, said Microsoft’s Exchange Online incident began at 16:17 UTC on Thursday, November 12, 2026. Outlook reports on the tracker began climbing at 17:25 UTC. By 18:25 UTC, reports about five Microsoft services combined reached 19,650 a minute, against fewer than 15 a minute on an average Thursday.

Outlook reports alone peaked at more than 2,000 times their normal level. The tracker counted 74 countries and territories in about an hour. Between 17:00 and 18:55 UTC, nearly 3 in 10 reports came from outside the United States.

THE OUTAGE IN FIGURES

  • User reports: Ookla counted more than 7 million outage reports by 16:00 UTC on Friday, November 13, the largest event on Downdetector in the past 12 months.
  • How long it stayed high: Reports ran high for about 20 hours, about twice as long as any other large outage in that window, and more than 90 countries or territories logged at least 1,000 reports.
  • Who pays for 365: Microsoft 365 has more than 450 million paid commercial seats. Microsoft has also said about 430 million people use the apps.
  • Two clocks: Consumer products recovered at 4:10 p.m. UTC on Friday, November 13, about 24 hours after the Exchange incident opened. Remaining customers were restored on Saturday, November 14, about 38 hours after that start.

The outage did not fade after the first peak. It rose again as each region started work on Friday, with busy hours around 9:00 a.m. in Japan, 9:30 a.m. in Singapore, 10:15 a.m. in India, and between 8:30 a.m. and 9:15 a.m. across much of Europe.

HOW THE OUTAGE UNFOLDED

  1. November 12, 2026, 16:17 UTC: Microsoft’s Exchange Online incident begins, later tracked as EX1550543 alongside the multi-service case MO1550821.
  2. November 12, 17:25 UTC: Outlook reports on Downdetector begin climbing, about an hour after the Exchange incident opened.
  3. November 12, 18:25 UTC: Combined reports on five Microsoft services hit 19,650 a minute.
  4. November 13, 2026: Asia, Australia and Europe log on to a still-broken work stack; Microsoft says the cause is a technical issue from an earlier network capacity upgrade that surfaced after a recent network change.
  5. November 13, 4:10 p.m. UTC: Microsoft says consumer products have recovered; enterprise 365 is still being brought back in stages.
  6. November 14, 2026: Microsoft 365 Status says the underlying issue is addressed and service is restored to remaining customers and regions.

Ookla’s comparison over 60-hour windows found this event drew about two and a half times as many reports as YouTube’s February 2026 outage, the next largest in the past 12 months. Most other large outages spike and fade within a few hours. The longest of those others stayed high for about nine hours.

Hospitals Went Back to Paper and Courts to Fax

The people who still got work done on Thursday and Friday were often the ones who still had a phone, a pad of paper, or a fax machine. That is a poor punchline for a productivity cloud with more than 450 million paid commercial seats, and it was the backup that actually worked.

In England, NHS staff turned to other IT systems or paper records under a five-year Microsoft 365 deal. The GMB union said ambulance services had suffered a serious impact. At Minshull Street Crown Court in Manchester, staff scrambled to find a fax machine. Some juries were sent home early. Some prisoners were not sent to court. Trades had to be confirmed by fax.

WHO FELL BACK ON OLDER TOOLS

  • Hospitals: NHS local services used other systems or paper records so appointments could continue.
  • Courts: Minshull Street hunted for a fax machine; staff elsewhere said judges could not open systems and prosecutors could not send papers.
  • Japan: Government email went down and staff used text messages; Japan’s Digital Agency could not receive email.
  • New Zealand: Immigration New Zealand paused visa processing.
  • United States: House staff lost a day of work, school districts told parents to use phone or other apps, and a Colorado detention facility said online bond payments were down.
  • Companies: Japan’s Taisho handled orders by phone; Sagawa Express could not take some online reservations; IndiGo and Akasa Air said email support was hit.

Australia’s ASIO and federal police were affected. ASIO director-general Mike Burgess said the agency’s systems operate on multiple levels and had been hit “on one level, yes, but not completely.” Service NSW asked people to delay visits. The US House, Japan’s government mail, and Immigration New Zealand all showed the same pattern: the office still existed, the inbox did not.

Teams was not a total blackout. Microsoft said joining calls, meetings, group chats and channel messages mostly still worked. Calendar scheduling, live events, Copilot replies that needed mail or files, and some private or shared channels did not. SharePoint file access stayed up even as external sharing and Exchange-tied features failed. Copilot Chat could not load, and it could not reach work email, calendars and files. Defender portals, Intune enrollment and Purview tools were in the same blast radius because they talk to Exchange.

Canberra Activated Its Microsoft Resilience Deal

Australia did not have a spare email system. It had a document. Home Affairs minister Tony Burke said the government activated a digital-resilience agreement with Microsoft so Australian interests would get priority as services came back.

“That agreement has in fact been activated today to make sure that we get appropriate prioritisation of Australian interests in this,” Burke said. Microsoft A/NZ president Jane Livesey and national cyber security coordinator Lieutenant General Michelle McGuinness briefed the government while ASIO and the Australian Federal Police were already on the affected list.

The prioritise recovery of Australia’s digital economy clause sits in a non-binding memorandum signed for Microsoft by Lisa Monaco, president of Microsoft Global Affairs. It tells both sides to work together after significant outages inside Microsoft’s own stack. It does not give Canberra another cloud. Asia-Pacific infrastructure was among the first regions Microsoft said it had remediated, which matches what a priority-restore ticket is for.

The same week showed how far the designation of “critical” had already gone. EU financial regulators named Microsoft a critical technology provider to banks and insurers in November 2025. In July 2026 the UK Treasury published new safeguards for major technology providers that put the same firm on the UK list. Those labels did not keep Outlook up. They describe what breaks when it falls.

A Similar Network Change Failed in October

This was not Microsoft’s first global failure built around a change that passed its own checks. The company’s October 2025 Azure Front Door review describes incompatible configuration metadata that moved through protection stages because health signals stayed green. The crash came minutes later. The last-known-good snapshot had already been overwritten with the bad config, so the safety net held the fault.

Impact then ran from 15:41 UTC on October 29, 2025, until 00:05 UTC on October 30, about eight hours and 24 minutes. Microsoft later blocked customer configuration changes, rebuilt the snapshot by hand, and rebalanced traffic slowly so recovering edge sites would not be overrun. The November 365 outage is a different product and a different bug. The shape is familiar: a change, a late failure, a restore that cannot be slammed back to full load.

TWO NETWORK FAILURES

Incident When What Microsoft described How long it ran
Microsoft 365 (MO1550821) November 12-14, 2026 Issue from an earlier network capacity upgrade, surfaced after a later network change Reports high about 20 hours; last customers restored Saturday
Azure Front Door (YKYN-BWZ) October 29-30, 2025 Incompatible configuration metadata that passed health checks, then crashed the data plane About 8 hours 24 minutes of customer impact
YouTube (Downdetector comparison) February 2026 Next-largest event on the tracker in the past 12 months Faded within a few hours; this 365 outage drew about 2.5 times as many reports

Mario De Cicco, vice president of European financial institutions at Morningstar DBRS, put possible global insured losses from this incident at $300 million to $500 million, below his range for the 2024 CrowdStrike outage. Switching costs still argue against a rush for the exits. The bill lands on lost hours, delayed lists and court days, not on a clean insurance cheque.

A Green Status Page Does Not Restore Lost Court Days

Anderson Economic Group, a Michigan research firm, estimated lost productivity could approach $7 billion. President Patrick Anderson said perhaps 26 million users on Thursday and 51 million on Friday saw disruption that caused an economic loss, and that many of the 430 million people who touch the apps were only annoyed. He called the figure a starting point issued before full restoration. Those two daily counts are not a combined headcount, because the same people can appear on both days.

Anderson also sketched a separate line: Microsoft could, in theory, owe about $800 million a day in unused subscription fees if customers pressed the point. Service credits in cloud contracts, where they exist, will not reopen a courtroom or restock a delayed order book. Australian Computer Society chief executive Prins Ralston said customers may be able to claim credits, and that those credits will not make up for a lost morning of trade for a small business.

Microsoft said it had not reported data loss and that delayed mail was being processed as queues cleared. The admin-center log for MO1550821 still told tenants it would watch the email backlog after the all-clear. A preliminary post-incident report is due within two business days of the close, and a final report within five. Until those papers appear, the record is an upgrade that introduced a fault, a later change that exposed it, and a Saturday restore that cannot give Thursday back.

Frequently Asked Questions

Which Microsoft 365 features kept working during the outage?

Microsoft said Teams core functions, including joining calls, meetings, group chats and channel messages, were mostly not hit, and that files stored directly in SharePoint Online remained reachable. Calendar invites, live or virtual events, Copilot answers that needed mail or files, and some private or shared channel actions failed because they depend on Exchange Online. Power Apps users already signed in on the web were told not to sign out, clear app data or reinstall while the issue ran.

When will Microsoft publish a post-incident report?

The MO1550821 tenant log said Microsoft will provide a preliminary post-incident report within two business days of the incident close and a final report within five business days. Charles Lamanna said the paper will include a step-by-step timeline and repair items. As of Saturday, November 14, 2026, that report had not been posted.

Did the outage cause other apps to look down too?

Ookla said GoDaddy, which resells Microsoft 365 email to small businesses, received more than 60 times its usual number of reports. Reports for Zoom rose more than 50 times, and reports for Gmail rose more than 40 times, in part because mail sent to Outlook users bounced and senders assumed the problem sat on their side. Xbox, GitHub and LinkedIn, also Microsoft properties, rose far less on the same tracker.

Can customers claim money back for the Microsoft 365 outage?

Cloud contracts often include service credits, and Prins Ralston of the Australian Computer Society said customers may be able to claim them, while warning that credits will not cover a lost morning of trade. Patrick Anderson of Anderson Economic Group modelled about $800 million a day in possible unused subscription fees if customers sought refunds, a separate line from his $7 billion productivity estimate and not a figure Microsoft has accepted.

Was email lost during the Microsoft 365 outage?

Microsoft has not reported data loss. Hexafusion’s incident note, drawing on Microsoft’s own guidance, said sending servers keep retrying, so queued mail should arrive as servers return, and that anyone missing an important message should ask the sender to send it again. The company said it would keep tracking the backlog after it restored remaining customers.

Harry is the editor of RTD JOURNAL, an independent publication that he owns, and ten years of journalism, first as a reporter, now as an editor, have left him with a habit of reading the documents other people skip. Annual reports are read to the footnotes, court filings to the exhibits, government releases to the methodology section, because that is where the numbers that matter usually sit. Each figure that reaches the page is checked against the document it came from, and claims that cannot be tied to a primary source are left out. That approach runs across the site's ten sections, written for an international readership: news, business and technology on one side, science, sports, entertainment, travel, lifestyle, gaming and auto on the other, all held to the same standard of evidence. A mistake, once found, is fixed on the article with a dated note that explains the change, as the site's public corrections policy requires. Readers can reach him with documents, questions or corrections at support@rtdjournal.com.

Click to comment

Leave a Reply

Your email address will not be published. Required fields are marked *

Trending