(The image is AI-generated)
Last updated: November 14, 2026
Microsoft 365 went down for much of the world on Thursday, November 12, 2026, and recovery stretched through Friday, November 13. Outlook, Teams and Copilot were among the services hit, and Downdetector reported problems in 74 countries and territories. Governments, hospitals and airlines fell back on phones, paper and fax.
Microsoft says this was a technical failure tied to a network capacity upgrade, not a cyberattack. That explanation is the company's own, and a full post-incident report has not been published yet. This article separates what is confirmed from what is still a claim, and it corrects several details in the earlier draft, including the date and the length of the outage.
- The outage began on Thursday, November 12, at around 1600 GMT, and ran for almost 24 hours before broad recovery, according to wire reports. It was not an 18-hour event that started on November 13.
- Microsoft attributes it to a fault from an earlier network capacity upgrade that surfaced after a recent network change, and says it was not cyber-related. The root-cause report is still pending.
- Downdetector counted reports from 74 countries and territories. That figure measures user complaints, not confirmed outages.
- Public bodies kept critical services running with workarounds. Immigration New Zealand paused all visa processing, and Unity Health Toronto declared a Code Grey.
- For defenders, the lesson is resilience: alternative communication channels, tested fallback plans and clear incident tracking.
Timeline: how the outage unfolded
Wire reports place the start at around 1600 GMT on Thursday, November 12. Downdetector logged problems in 74 countries and territories within about an hour. At its peak, the site showed more than 130,000 Outlook problem reports in the United States, according to Reuters. Sources give slightly different start times for early user complaints, so treat the exact minute as approximate.
Microsoft first said it was investigating reports in the Americas, then confirmed that Microsoft 365 connectivity could be affected for users in any region. Engineers restarted servers, but the company said servers brought back online could become unresponsive again once traffic returned. At one point it said a change it had applied was not taking effect as expected.
By Friday morning, Microsoft reported partial recovery across regions. In the afternoon, reports described service as broadly recovering nearly 24 hours after the first problems, with email queues approaching normal levels. Microsoft said it had no ETA for full recovery at that stage, and it kept monitoring the backlog of delayed email.
What Microsoft says caused it
Microsoft's public explanation, posted on X on November 13, was short. It called the event a technical issue unrelated to cyber activity or AI agents, introduced during an earlier network capacity upgrade and surfacing after a recent network change. It said it had addressed the underlying issue and was restoring remaining customers and regions.
Earlier status messages gave more mechanics. Microsoft said the affected infrastructure was under elevated load across regions. Reporting by RTE says the company described that load as coming from servers becoming unexpectedly unavailable after failing automated health checks, not from a surge in traffic. A status update quoted by TechRadar said a core process on the affected servers was becoming overwhelmed and stopping responses, which reduced capacity.
These pieces fit together, but they are all Microsoft's account. No independent party has verified the root cause. Microsoft executive Charles Lamanna said a post-mortem with a step-by-step timeline and repair items would come shortly.
Outside experts offered cautious readings. Brazilian analyst Thiago Bordini said a configuration error in interconnected environments can spread and overload servers, which is plausible but not confirmed for this event. Australian minister Tony Burke said Microsoft had likened the incident to the 2024 CrowdStrike failure, meaning an update that went wrong. The mechanism differs: CrowdStrike's faulty update crashed Windows machines, while this event involved Microsoft's own cloud services.
Which services were affected
Microsoft's service health notices listed Exchange Online, Microsoft Teams, SharePoint Online, Microsoft 365 Copilot and Microsoft Defender XDR, among others. Reports also named Planner, To Do and Intune. Administrators tracked the incident in the Microsoft 365 admin center under MO1550821, with EX1550543 for Exchange Online. Consumer services such as Xbox sign-ins were also reported as affected.
Whether Azure was hit is unclear. Some reports list Azure among affected services, while Microsoft's own list centered on Microsoft 365 workloads. I could not confirm either version from a primary source.
CBC and Time reported that OpenAI and Anthropic users saw failures involving Microsoft 365 connectors and integrations. That is a narrower problem than a general outage of those AI services. Anthropic develops Claude, the assistant used to prepare this text, and the account here comes from the media reports, not from Anthropic.
Impact on governments, health care and transport
Government
In the UK, cyber minister Liz Lloyd said the impact on critical frontline services such as 999 calls was limited and that transport was not disrupted. Some police forces and schools were affected. The Singapore government said email was down. The Philippines Supreme Court said some operations, including email responses, were affected. In Japan, Deputy Chief Cabinet Secretary Masanao Ozaki said government IT infrastructure was affected, and officials arranged alternatives such as telephones.
In Australia, the domestic spy agency and federal police were affected. In New Zealand, the foreign ministry lost email with overseas posts, and Immigration New Zealand paused all visa applications. It said normal processing would resume on Monday.
Health care
NHS England said staff used paper records or other IT systems and told patients to attend appointments unless told otherwise. In Canada, Unity Health Toronto declared a Code Grey, the alert used for infrastructure loss, and said patient care was not affected.
Transport and business
Indian carriers IndiGo, Akasa Air and Air India said their email support services were affected. Auckland Transport said it lost access to Outlook, SharePoint and Teams. Traders in Japan reportedly went back to fax to confirm deals. These are examples drawn from news reports, not a full census of affected organizations.
Microsoft's response
On Friday Microsoft said it was deeply sorry and that it was seeing broad recovery and stability. It also said it had restored services gradually, validating each stage, which took longer but helped keep recovery stable. It repeated that it was not aware of malicious activity and promised a full post-incident report with no date attached.
Analyst Carmi Levy told CBC that even after an all-clear, the message backlog could take hours or days to clear, and that experiences would vary by user. This matters for planning: an outage ends in the status page before it ends in your inbox.
What it says about cloud dependency
Microsoft said in January that it had more than 450 million paid commercial Microsoft 365 subscriptions. That is Microsoft's own figure and does not measure how many people were affected here. It does show why a fault in a shared layer reaches public services in many countries at once.
This was the second Microsoft 365 disruption of 2026, after an outage in January, and CBC also noted an Azure outage last month. Attacks leave different traces from this kind of failure. For comparison, see our coverage of the ShinyHunters jobs portal breach and the Binance hack, where intruders were involved, and our roundup The September Storm of attacks from September 10 to 24.
Brazilian lawyer Ronaldo Lemos argued for a digital resilience policy that maps dependencies on specific services and keeps a Plan B, C and D. Rob Demain, chief executive of e2e-assure, called the event a reminder of reliance on a handful of big providers. Both are opinions, not findings.
Claim check
This table covers claims in the earlier draft and in early coverage. Confirmed means reliable reporting supports it. Company claim means Microsoft or another party asserts it and I found no independent check. Unverified means I found no solid source. Wrong means current sources contradict it.
| Claim | Status | Note |
|---|---|---|
| The outage was a November 13 event. | Wrong | It began Thursday, November 12. November 13 was the second day. |
| It lasted about 18 hours. | Wrong | Reports describe almost 24 hours before broad recovery. Microsoft has not published an official duration. |
| The first spike was at 6:15 PM GMT on Teams in the UK. | Unverified | Wire reports say the outage began around 1600 GMT. Removed. |
| Users in 74 countries and territories reported problems. | Confirmed | Downdetector's count, reported by several outlets. It counts user reports. |
| More than 130,000 US Outlook reports. | Confirmed | At the peak, according to Reuters. |
| A network capacity upgrade fault surfaced after a network change. | Company claim | Microsoft's explanation. Root-cause report pending. |
| It was not a cyberattack. | Company claim | Stated by Microsoft and relayed by Australia's Tony Burke. No independent forensic review published. |
| Restarted servers became unresponsive again under traffic. | Confirmed | Microsoft's own status message, reported by wire services. |
| Microsoft declared services stable after 18 hours. | Unverified | Microsoft reported broad recovery and was still watching the email backlog. I found no formal all-clear. |
| Azure was not affected. | Unverified | Reports conflict. Removed. |
| ChatGPT and Claude were disrupted because they depend on Microsoft infrastructure. | Wrong | Reports describe failures in Microsoft 365 integrations. The infrastructure dependency was speculation. |
| UK minister quotes and a call with Microsoft UK's chief executive. | Unverified | Reported: limited impact on critical frontline services. Exact quotes and the call not found. Removed. |
| Japan's official was Masayoshi Ozaki and the system was GSS. | Wrong | The official is Masanao Ozaki. Reports say government IT infrastructure, with no system name. |
| Burke said Microsoft likened it to CrowdStrike. | Confirmed | He described an update that went awry. Reports vary on his exact title. |
| It rivals CrowdStrike in scale. | Unverified | One Oxford professor said it feels like it could be worse. No loss figures published. |
| A professor called the 18-hour duration unprecedented. | Unverified | Quote not found, and it relied on the wrong duration. Removed. |
| Bordini and Lemos comments. | Confirmed | Reported in Portuguese. Paraphrased here. |
| Malaysia Airlines, Emirates, National Rail and named banks were affected. | Unverified | Not found in sources I could reach. IndiGo, Akasa Air and Air India are confirmed. Removed the rest. |
| NHS reverted to paper records. | Confirmed | NHS England said paper records or other IT systems. The exact earlier quote was not found. |
| It was the second Microsoft 365 disruption of 2026. | Confirmed | After the January incident, per Reuters and CBC. |
Update
Change log: November 14, 2026 – rewritten and fact-checked; corrected the start date, duration and several names and quotes. November 14, 2026 – first published.
My take
- An outage this large can cause serious problems worldwide, and it should push companies like Microsoft to treat security and resilience as a first-order responsibility. Health services, courts and immigration offices were forced into workarounds because one provider's network had a bad day.
- One reservation: this was not an attack, according to Microsoft, so the right lesson is about change management and availability as much as defense against intruders. Microsoft's cause is still unverified, and the post-incident report will show whether this is a one-off or a pattern after the January outage.
- I have not tested any of these systems myself. This view comes from reading wire reports, official statements and expert comments.
What this means for you
- Keep a second channel. Maintain phone trees or a chat tool outside Microsoft 365 that staff can reach when email and Teams fail.
- Test the fallback. Practice paper or manual procedures for critical processes before you need them.
- Map your dependencies. List which services rely on Exchange Online, Teams or SharePoint, including integrations and connectors.
- Track the incident properly. Admins can follow MO1550821 and EX1550543 in the Microsoft 365 admin center under Service health.
- Plan for the backlog. Expect delayed mail after recovery, and tell users that delivery can lag the status page.
- Do not assume an attack, or rule one out. Check for signs of compromise in your own logs while waiting for the vendor's report.
FAQ
Was the Microsoft outage a cyberattack?
Microsoft says no, calling it a technical issue unrelated to cyber activity. That is the company's statement, and no independent investigation has been published.
How long did it last?
Reports describe almost 24 hours from the first problems on Thursday to broad recovery on Friday afternoon. Microsoft has not given an official duration, and some users saw delays after that.
When will Microsoft publish the full report?
Microsoft has promised a post-incident report without a date. An executive said a post-mortem would come shortly. Check Microsoft's service health channels for it.
Sources
- Microsoft 365 Status on X: recovery update, November 13, 2026
- Windows Central: Microsoft apologizes for major outage (reproduces Microsoft's statement)
- Yahoo News (wire reports): Microsoft struggles to fix global outage; Microsoft deeply sorry after almost 24 hours
- ABC News (Australia): Microsoft says sorry for global outage
- CBC News: No timeline for full fix; Microsoft reports progress
- CBS News: Continued recovery progress; NBC News: Microsoft scrambles to fix major outage; TIME: Microsoft struggles to repair outage
- RTE: What went wrong and has it been fixed?; TechRadar: Live coverage; Engadget: Multiple services outage
- The Jerusalem Post: Recovery after outage lasting over 24 hours (Reuters-based figures)
- Stuff (New Zealand): Outage downs systems around New Zealand
- Já é Notícia (Portuguese): Bordini and Lemos comments; Olhar Digital: 74 countries
- Cybersecurity News: Incident IDs and affected services
