Microsoft 365 outage passes 140,000 reports across 74 countries and stretches into day two

What happened

Microsoft 365 has been partially offline since Thursday, November 12, and the disruption was still unresolved on Friday morning. According to a dev.to write-up that compiles Downdetector figures via BBC and People reporting, outage reports peaked at more than 140,000 Outlook users spread across 74 countries.

The trouble began shortly before 11:17 a.m. ET, when Exchange Online customers in the Americas could no longer reach their mailboxes. Within roughly two hours, per the dev.to account, Microsoft escalated the problem to a global incident tracked under two identifiers: EX1550543 for Exchange Online and MO1550821 for Microsoft 365 as a whole. By 1:25 p.m. ET, Downdetector was logging nearly 20,000 reports per minute across five Microsoft services, compared with the trickle of a normal Thursday.

Nearly the entire product family was degraded: Outlook, Exchange Online, Teams, SharePoint Online, Microsoft 365 Copilot, Defender XDR, Planner, To Do, Intune, Purview and Power BI. Public-sector bodies were caught up too, with the dev.to summary naming the U.S. House of Representatives, the UK's NHS, police forces, schools and Parliament among the affected organizations.

Why the fix keeps failing

The unusual part of this incident is the recovery itself. Microsoft's status updates, relayed by Cyber Security News and the dev.to post, concede that servers brought back online may stop responding again once real user traffic hits them. Microsoft is therefore throttling how quickly it reintroduces traffic while separate workstreams try to pin down what triggers the behavior.

The dev.to author reads this as a textbook cascading failure, a variant of the thundering-herd problem: when a large slice of capacity disappears, the remaining servers inherit the entire load, more of them fail, and every restarted machine is handed the same impossible demand the moment it returns. The remedy is not more restarts but load shedding until capacity and demand rebalance — which is why Microsoft is easing traffic back rather than switching everything on at once.

Microsoft's public explanation so far is thin: a large number of servers went dark without warning, and the leftover infrastructure was overwhelmed by the redirected load. Executive Charles Lamanna said on X that there was no sign of malicious activity, and Microsoft has promised a full post-incident report with a root cause once mitigation finishes. As of Friday morning, there was neither a root cause nor an estimated time to full restoration.

LibreOffice takes the opening

Midway through the outage, LibreOffice posted a dig playing on Microsoft's branding, noting that its own suite runs all year round. The dev.to author ran a quick check: LibreOffice 24.2.7.2 converted a document to PDF locally in about three seconds, without touching an account, a server or the network.

The comparison only goes so far, as the author concedes. A desktop office suite cannot resurrect a dead mailbox, a failed chat platform or an unreachable AI assistant. The real lesson, the piece argues, is about blast radius: when mail, collaboration, files and AI features all depend on one control plane, a single bad day takes them down together — while the document being edited is the one layer that never needed to be a cloud service. With Microsoft 365 Personal listed at $99.99 a year in the US, it is also a cheap moment to ask whether any local fallback exists at all.

Why it matters

Two things decide how this story ends. First, whether the post-incident report names an actual trigger — Microsoft's current wording describes a symptom, and a bad configuration push, a faulty health check and a capacity-planning miss imply very different levels of confidence in the fix. Second, whether the throttled recovery holds or restored servers keep flapping; another day of unstable service would shift this from a service outage to a question about capacity planning. For the businesses, hospitals and legislatures that route email, meetings, files and AI assistance through a single vendor, the summary is already stark: 140,000 reports across 74 countries is what one control-plane failure looks like.