Microsoft 365 goes down for hours as Exchange servers fail health checks
Microsoft's cloud productivity stack failed on Thursday, with Exchange Online servers failing automated health checks and dropping out, overloading the servers still running. The fault spread to Outlook, Teams, SharePoint Online and Microsoft 365 Copilot, with Outlook problem reports peaking at 138,058 around 1 p.m. ET and reports reaching 74 countries and territories within about an hour.
The News
Microsoft's cloud productivity stack failed on Thursday in an outage that ran for hours. Exchange Online servers failed automated health checks and dropped out of service, overloading the servers still running; the fault spread to Outlook, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Intune and Purview. In plain terms, the email system broke first, and the extra load it pushed onto the remaining servers broke much of the rest.
Reports to Downdetector about Outlook began climbing around 12:25 p.m. ET; Microsoft acknowledged the problem at 12:54 p.m. ET and tracked it as incidents EX1550543, for Exchange Online, and MO1550821, for the wider Microsoft 365 suite. Outlook problem reports peaked at 138,058 around 1 p.m. ET and had fallen to about 34,490 by evening, while outage reports reached 74 countries and territories within about an hour. The outage also hit connectors used by OpenAI's and Anthropic's AI services.
Microsoft said it did not suspect malicious activity, that a small number of configuration updates were under investigation as potential contributors, and that a core process on the affected servers was becoming overwhelmed. An alternative health-check method let some servers return from about 4 p.m. ET, with gradual recovery after that. How long the outage lasted is reported as a range — more than five hours by Reuters' early version, about ten hours by its later version. Microsoft's status updates pointed to access being restored around 05:33 UTC on Friday, although Downdetector still showed more than 25,000 Outlook problem reports at 02:30 UTC on Friday and Microsoft's own status page still listed service degradation earlier that morning. No final root cause has been published. Microsoft shares rose 0.36% on the day.
Email broke first — then the overload broke almost everything else.
Why It Matters
The failure mode is what makes this outage worth understanding. Exchange Online servers run automated health checks — self-tests that confirm a server is working properly. When those checks failed, the servers dropped out of service, and the work they had been doing landed on the servers still running. Those servers then became overwhelmed, and the fault spread outward to Outlook, Teams, SharePoint Online, Copilot and more. It is a cascade: one system's failure creates the overload that breaks the next one, which is why an email problem became a whole-productivity-stack problem within the hour.
The scale shows how central Microsoft 365 has become to daily work. Reports reached 74 countries and territories within about an hour, and Outlook problem reports peaked at 138,058 around 1 p.m. ET. Those are Downdetector reports — user complaints, not a count of affected people — but they are a useful measure of how fast and how widely the disruption was felt. The outage also reached the connectors used by OpenAI's and Anthropic's AI services, a reminder that Microsoft's stack is now infrastructure for other companies' products too, not just its own.
The market's reaction was calm, and that is informative. Microsoft shares rose 0.36% on the day — the market treated it as an operational failure, not a demand signal. Nobody stops paying for email because it was down for a day, and switching costs for a business running on Microsoft 365 are enormous. But calm is not the same as forgiving: reliability is the product here, and a failure that starts in a core process becoming overwhelmed, with a small number of configuration updates under investigation, is the kind of thing customers remember the next time a contract comes up for renewal.
The unsettled parts are the duration and the cause. Reuters' own versions differ — more than five hours in the early version, about ten hours in the later one — and one tracker still showed degradation at 04:10 UTC on Friday even though Microsoft reported access restored around 05:33 UTC. No final root cause has been published. Until Microsoft explains what happened and what changes, the story is incomplete, and the honest reporting is a range, not a single settled number.
Analyst's Take
Our call is Mixed. The outage was serious — a core email system failing its own health checks, cascading across Outlook, Teams, SharePoint and Copilot, in 74 countries, for hours — and the absence of a final root cause leaves the obvious question unanswered. But the market treated it as an operational failure, not a demand signal: Microsoft shares rose 0.36% on the day, and there is no evidence customers are leaving. Reliability is Microsoft's core promise, and this dented it, but one bad day does not undo a decade of uptime.
What to watch: the root-cause report. If Microsoft publishes a clear explanation and a credible fix — and if the alternative health-check method that brought servers back from about 4 p.m. ET becomes permanent — this fades into an embarrassing footnote. If the root cause stays unpublished, or if a second outage follows, customers will start asking harder questions, and the market's calm will not last.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is Exchange Online?
Exchange Online is Microsoft's cloud email service — the part of Microsoft 365 that handles sending, receiving and storing email for businesses. When its servers failed their automated health checks and dropped out, email stopped working, and the overload on the remaining servers spread the failure to Outlook, Teams, SharePoint Online and the rest of the Microsoft 365 stack.
What is Downdetector, and what do its numbers mean?
Downdetector is a website that tracks outage reports from users — people go there to say a service is not working. The numbers are reports, not a count of affected users: Outlook problem reports peaked at 138,058 around 1 p.m. ET and had fallen to about 34,490 by evening. They show the scale and timing of the disruption, not an exact headcount.
Was it a cyberattack?
Microsoft said it did not suspect malicious activity. A small number of configuration updates were under investigation as potential contributors, and the company said a core process on the affected servers was becoming overwhelmed. No final root cause has been published, so the exact trigger is still not confirmed.
How long did the outage last?
That is not settled. Reuters' early version reported more than five hours of disruption on Thursday, while its later version reported about ten hours. Microsoft reported access restored around 05:33 UTC on Friday, and one tracker still showed service degradation at 04:10 UTC on Friday. The honest answer is a range, not a single number.
What should users and investors watch for now?
Whether Microsoft publishes a final root cause, and whether it explains what changes so the same failure cannot repeat. For users, the question is whether the alternative health-check method that let some servers return from about 4 p.m. ET becomes a permanent fix. For investors, the test is whether customers treat this as a one-off or start asking harder questions about reliability.
👁
Source: TBS News