Tech • AI • Robotics • Game

ENFR

Why Microsoft couldn’t fix its own servers for 36 hours...

7/10
AIFireshipNovember 16, 2026 at 08:18 PM6:22
Audio player

TL;DR

A major Microsoft 365 outage disrupted Outlook, Exchange Online, SharePoint, Copilot and other services across dozens of countries for about 36 hours, after a network capacity change triggered cascading server failures.

KEY POINTS

Global Microsoft 365 disruption

The incident began at about 11:17 a.m. Eastern on Thursday, when Exchange Online mailboxes stopped opening. Reports quickly spread across 74 countries, and the outage affected not only Outlook email but also a wider set of Microsoft 365 services, including SharePoint and Copilot. The disruption lasted roughly 36 hours, with service restoration stretching into Friday.

What failed inside Microsoft

Microsoft said the problem was not caused by unusual traffic volume but by a large number of servers becoming unexpectedly unavailable. The company later tied the failure to an earlier network capacity upgrade and said routine maintenance exposed a latent bug in the process that establishes secure connections. That caused connection requests to pile up until servers ran out of processing power.

Cascade driven by health checks

The outage appears to have spread through a classic load-balancing failure. When servers failed automated health checks, they were removed from rotation, shifting traffic onto remaining machines. Those surviving servers then became overloaded, failed their own health checks, and were pulled out as well, creating a domino effect across the platform.

Why restarting servers did not solve it

Recovery was slowed because bringing servers back online did not remove the underlying pressure. As soon as restored systems received traffic, they risked becoming unresponsive again because clients worldwide were repeatedly trying to reconnect. Microsoft also used an alternative health-check approach during mitigation, effectively testing whether servers could handle traffic despite failing standard health signals.

Admin tools were hit too

The outage also disrupted the Microsoft 365 admin center, limiting customers’ ability to investigate service health or manage the incident from their side. That compounded the operational impact for businesses and public agencies already dealing with lost email access, delayed workflows and service interruptions across connected Microsoft products.

Real-world fallout

The disruption forced some organizations into manual workarounds. Government offices reportedly returned to paper processes in some cases, and one Colorado jail had to accept bond payments in person. Financial firms on Wall Street also reverted to more traditional communications tools as email service remained unstable.

FAA shutdown was separate

On Friday, the Federal Aviation Administration separately halted U.S. flight departures for roughly five hours, leading to more than 9,000 flight cancellations. Officials said the aviation incident was not linked to Microsoft. Homeland Security and Microsoft both stated there was no evidence connecting the two failures.

Preliminary FAA explanation

Transportation officials indicated the FAA issue involved overloaded switches and routers after a router replacement at a remote site. One account said the change triggered around 1,000 login attempts every four to five seconds, flooding other network equipment. Other reports pointed to aging Cisco infrastructure, but no final FAA root-cause report had been released.

Shared lesson across both outages

Although the two incidents were separate, they highlighted a similar weakness: changes to critical network infrastructure can overwhelm core systems that many other services depend on. In Microsoft’s case, a capacity-related network change appears to have reduced resilience instead of improving it, turning a maintenance event into a broad service failure.

CONCLUSION

The outage showed how a single network change inside a major cloud platform can ripple across email, collaboration and administrative systems worldwide. Microsoft’s final post-mortem is still awaited, but the preliminary picture points to a preventable cascading failure in core service infrastructure.

Ask a question

More from AI