Microsoft 365 Outage Takes Down Mail and Copilot, Undelivered Messages Held for 48 Hours
Since Thursday, Microsoft 365 mail, cloud services and Copilot have been failing: servers fail the automated health check that asks "are you alive?" and drop out of service. Machines brought back online hang again as soon as they receive traffic. That is why Microsoft extended the retention period for undelivered email to 48 hours, so messages wait in a queue rather than disappear. The takeaway for teams: don't keep login codes on the same email provider.
AI-processed from Ars Technica; edited by Hamidun News
On Thursday afternoon, a global outage hit Microsoft 365, and it still hadn't ended by Friday morning. Outlook mail, cloud services and the Copilot AI assistant all stopped working reliably. According to Microsoft, the cause is that servers are failing the automated health check in large numbers. That is a routine request asking "are you alive?", and when servers fail it, they drop out of service. To keep emails from being lost while mail is down, the company extended how long messages stay queued to 48 hours. If your team relies on Outlook, Teams and cloud documents, you have most likely felt this already.
What broke and why the fix is taking so long
Microsoft said the problem began in a part of the infrastructure that was under heavy load in several regions at once. The company stresses that the cause is not a rise in traffic, but that servers are failing automated checks and becoming unavailable.
Here is how it works in a simple example. A load balancer sits in front of a group of servers and distributes user requests. It constantly asks each server, "are you OK?" If there is no answer, the machine is removed from rotation. When the checks break on many servers at once, the load falls on the remaining ones, they become overloaded, and they stop responding too. The result is a domino effect.
Microsoft first tried to stop new servers from dropping out and checked other ways of running the health check. After four hours, the company admitted that the measures it had taken "did not provide the expected relief." It turned out that a basic process was overloading on the affected servers, which reduced available capacity. Engineers then began isolating part of the infrastructure, restarting components and gradually returning traffic while checking that services stayed stable.
But even that did not work right away: some servers that were brought back online stopped responding again as soon as requests reached them. So Microsoft moved to protect data: the hold period for undelivered mail was increased to 48 hours. Put simply, if an email cannot be delivered immediately, the system waits up to two days and does not discard it.
Who was affected
Ciaran Martin, former head of the UK's National Cyber Security Centre, told NBC News that the outage was one of the largest in history and suggested it could turn out to be worse than the recent failures at Amazon Web Services and CrowdStrike. This is an expert's assessment; there is no final tally of the damage yet.
NBC describes the following consequences:
- Japan's government IT infrastructure was disrupted, said Deputy Chief Cabinet Secretary Masanao Ozaki;
- doctors could not send patients important messages about medication changes and further treatment;
- people could not make bank transfers because confirmation codes did not arrive;
- the Supreme Court of the Philippines could not conduct its legal work;
- one sheriff's office in Colorado had difficulty processing bail for immediate release of detainees.
It is unknown exactly how many users have been affected. For scale: in January, Microsoft put the number of paid commercial Microsoft 365 subscriptions at more than 450 million.
What is known about recovery and what to do
By Friday, Microsoft reported improvements in North America, stable recovery across most of Europe and "encouraging signs" in the Asia-Pacific region. According to DownDetector, the peak of complaints came on Thursday afternoon and had fallen noticeably by Friday morning. However, one Outlook user writes that outgoing emails go through normally while incoming ones arrive unreliably. Some Copilot users complain that the AI assistant came back later than mail.
Microsoft says it sees no signs of malicious activity and promises to publish a full post-incident report covering the root cause and protective measures once the outage is resolved. Until that report is out, it is too early to judge why the health checks broke in the first place.
What teams building products should do:
- Do not keep two-factor authentication codes and important customer notifications on a single mail provider; have a backup channel, such as SMS or a messenger;
- decide in advance what your AI features do if the vendor's cloud is unavailable: a manual mode or a fallback model;
- review your own health checks: if they remove servers from rotation too aggressively, a partial failure can turn into a cascade;
- prepare a customer communication plan for when both mail and internal chat are down.
The main lesson of this outage is that a mass failure cannot be fixed by a simple restart. Servers come back, take on real load and fail again. Conclusions about causes and damage should wait for Microsoft's report.
- Nov 21, 2026Microsoft outage: nearly 83,000 Outlook complaints on Downdetector, Teams and Copilot down too
- Nov 21, 2026MLPerf Training Adds Nvidia Rubin and TPU Ironwood in First Appearance: 153 Systems
- Nov 21, 2026Microsoft Turns Copilot Into an "Operating System for Work": Word and Excel Become Agent Plugins
- Nov 21, 2026Google to split Gemini chat history from model training, so you can opt out without losing your history
Key Questions Answered
What broke and why the fix is taking so long?
Microsoft said the problem began in a part of the infrastructure that was under heavy load in several regions at once. The company stresses that the cause is not a rise in traffic, but that servers are failing automated checks and becoming unavailable. Here is how it works in a simple example. A load balancer sits in front of a group of servers and distributes user requests. It constantly asks each server, "are you OK?" If there is no answer, the machine is removed from rotation.
Who was affected?
Ciaran Martin, former head of the UK's National Cyber Security Centre, told NBC News that the outage was one of the largest in history and suggested it could turn out to be worse than the recent failures at Amazon Web Services and CrowdStrike. This is an expert's assessment; there is no final tally of the damage yet.
What is known about recovery and what to do?
By Friday, Microsoft reported improvements in North America, stable recovery across most of Europe and "encouraging signs" in the Asia-Pacific region. According to DownDetector, the peak of complaints came on Thursday afternoon and had fallen noticeably by Friday morning. However, one Outlook user writes that outgoing emails go through normally while incoming ones arrive unreliably. Some Copilot users complain that the AI assistant came back later than mail.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.