Microsoft 365, Outlook and Teams outage passed five hours Thursday

Microsoft's Microsoft 365, Outlook and Teams services stayed disrupted for thousands of users for more than five hours on Thursday, and the outage also broke connectors relied on by OpenAI's and Anthropic's AI services.

As of 6 p.m. ET, Microsoft was still working to bring affected servers back into service. The company said it had observed limited recovery as users gradually reconnected. It's a strikingly long window for a productivity suite that businesses, schools and now AI vendors treat as infrastructure.

How bad did the outage get?

The numbers tell the story. Outlook problem reports on Downdetector peaked at 138,058 around 1 p.m. ET on Thursday, then fell to roughly 34,490 by evening. By 5:47 p.m. ET, Downdetector showed nearly 13,000 reports of Microsoft 365 problems and 763 reports of trouble accessing Teams.

The scale matters. Microsoft 365 sits at the center of corporate email, calendars and chat for a large share of the global workforce, so a multi-hour Exchange Online disruption translates directly into delayed business operations, missed messages and postponed decisions. When a platform that size wobbles, the ripple effects reach companies that never expected to depend on Microsoft's uptime.

A Microsoft spokesperson acknowledged the problem directly: "We are aware and investigating reports of users experiencing issues with Microsoft 365 services, primarily issues with Exchange Online."

What caused the disruption?

Microsoft pointed to a server-side failure, not a surge in demand. The company said earlier in the day that a number of servers had become unexpectedly unavailable under elevated load — and, importantly, that the problem was not caused by an increase in traffic.

That distinction carries weight for engineers and customers alike. A capacity shortfall caused by unexpected demand suggests one kind of fix. Servers dropping out under load they should normally handle points to a different failure mode, one that raises questions about resilience and failover in Microsoft's cloud architecture.

Later in the day, Microsoft offered a cautiously positive update. The company said the servers it had restored so far remained stable, and that it was shifting traffic to alternate routes wherever possible while streamlining the process of bringing the remaining impacted servers back online.

In a statement, Microsoft set expectations for how the fix would look from the user side: "As capacity is restored, users may experience a gradual recovery as their clients reconnect."

Gradual is the operative word. Falling Downdetector counts don't mean full service — they mean fewer people hitting a wall at the same moment.

How did the outage hit AI services?

The disruption didn't stop at email and chat. It reached into the AI industry's own plumbing.

OpenAI said it was investigating an issue with Microsoft connectors in which new connections may be broken and tools were not working as expected. Anthropic said it had identified an issue with an upstream provider affecting its Microsoft 365 connector.

Both statements point to the same structural reality. AI assistants that read mailboxes, calendars and documents through Microsoft 365 connectors inherit Microsoft's availability. When Exchange Online falters, the AI tools built on top of it falter too — and users of ChatGPT or Claude integrations experience an OpenAI or Anthropic outage that neither company controls.

What was the business fallout?

One company moved financial disclosures because of it. Electric truck maker Workhorse Group, which was scheduled to report third-quarter results on November 12, rescheduled the announcement to November 16, citing business disruptions caused by the global Microsoft outage.

A publicly traded company postponing earnings by four days is a concrete measure of real-world cost. Earnings dates rarely move for operational reasons, and Workhorse's decision signals that the outage disrupted internal processes — reporting, review or communication workflows — badly enough to delay a regulated disclosure.

Is this part of a pattern?

Thursday's incident follows another significant failure. Microsoft reported an outage of its Azure cloud platform just last month, and that earlier outage affected the company's productivity software suite along with a range of industries worldwide.

Two impactful disruptions in consecutive months put pressure on Microsoft's reliability record. For enterprises weighing where to run critical workloads, and for AI companies building products on Microsoft connectors, back-to-back outages turn reliability from a background assumption into an active procurement question.

What happens next?

Microsoft's stated path forward is traffic rerouting and a staged server recovery, with users reconnecting gradually as capacity returns. The company said restored servers had held stable so far.

The broader question is whether Thursday's server failures under elevated load prompt a public root-cause explanation. With Workhorse rescheduling earnings, OpenAI and Anthropic acknowledging broken connectors, and Outlook reports peaking above 138,000, customers across the stack — from corporate IT desks to AI product teams — now have fresh reasons to ask Microsoft what failed, and what changes before the next outage.