Microsoft 365 back online after 18-hour outage: Microsoft blames network upgrade

Microsoft 365 back online after about 18 hours of disruption: on Friday afternoon (13), Microsoft said it has addressed the problem behind Thursday's (12) global outage and is seeing broad recovery and stability across its services. The incident, however, has not been declared closed — the company is still restoring access for the remaining customers and regions.

What is back, and what is still missing

By Friday morning, the backlogs of emails accumulated during the outage were already close to normal levels, according to the company. Mailbox access returned for most users, but Microsoft warned that intermittent access failures and delayed deliveries may still occur while the restoration is completed.

In practice, getting into Outlook does not guarantee that every message has arrived: part of the queued mail was still being processed on Friday afternoon. To protect messages in transit, Microsoft temporarily extended the delivery timeout window to 48 hours — which does not mean services will take 48 hours to return, nor does it guarantee delivery of every message within that window.

  • Outlook, Teams, SharePoint Online and Copilot: access broadly recovering, with residual impact in some regions.
  • Exchange Online: accumulated message queues almost back to normal on Friday morning.
  • AI connectors: OpenAI and Anthropic reported problems with their Microsoft 365 connectors during the outage.

The new diagnosis: a network capacity upgrade

The main news on Friday is Microsoft's first explanation of the cause. In a post on the company's official X account, Microsoft apologized and said: "This was a technical issue not related to cyber or AI / agents. The incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change. We have addressed the underlying issue and continue to restore service to remaining customers and regions."

The company also promised a full post-incident report once its review is complete, with no date yet. Microsoft said it restored services gradually, validating stability at each stage — which, according to the company, deliberately prolonged the return.

The explanation changed three times in under 24 hours

The diagnosis released on Friday is the company's third version of what caused the outage:

  1. On Thursday, Microsoft blamed elevated load caused by servers becoming unexpectedly unavailable after failing automated health checks — and said traffic had not even increased and that it had found no recent change linked to the start of the problem.
  2. Overnight, the company began talking about a core process responsible for establishing secure connections that became overwhelmed, and started rolling back a recent network configuration change.
  3. On Friday afternoon, the latest version: the fault came from an earlier network capacity upgrade, exposed by a recent change.

What happened on Thursday

The outage began on Thursday afternoon (12), at around 13:17 Brasília time, in Exchange Online, and spread to the services that depend on it. Outlook problem reports on Downdetector reached around 138,000. Geekiki's initial coverage of the outage: Microsoft 365 outage leaves Outlook, Teams and Copilot down for hours.

According to Kentik, a company that monitors internet traffic, traffic returned to Microsoft's services on Friday morning after about 18 hours down — and it is the second time in under three months that Microsoft 365 has been down for hours at a time.

Why it matters

The episode shows how a failure in a single internal network component can take down email, collaboration and AI assistants at the same time — and how the diagnosis of an outage this large can change several times while the investigation continues. Until the full report is published, administrators are advised to follow incidents EX1550543 and MO1550821 in the Microsoft 365 admin center.

Sources

Featured image: Outlook icon (Microsoft, via Wikimedia Commons, CC BY-SA 4.0).

This content was produced with the assistance of artificial intelligence and reviewed editorially.

By Muse