Field report N-51 / M9 · sheet /operations

Operations & RestructuringFiled from the territory

Microsoft Traces Global Outage to Bug Hiding Since July

A gateway bug shipped in July knocked out over 80% of Microsoft 365 traffic servers after a Teams reroute, hitting 74 countries and disrupting governments before fixes landed.

Grid
N-51 / M9
Surveyed
November 20, 2026
Reading
3 min read
Correspondent
Elena Vasquez
Microsoft traced its global outage to a software bug lurking since July
Fig. N-51 / M9Microsoft traced its global outage to a software bug lurking since July — Innovalabs / pixabay

Field notes

  1. Bug lived in Microsoft 365 gateway servers since July 2024; load tests on weaker lab hardware missed it.
  2. Over 80% of gateway servers stopped responding within two hours of the Nov. 12 incident.
  3. Outage ran from 11:20 a.m. ET Nov. 12 to 11:10 a.m. ET Nov. 13, with reports from 74 countries.
  4. Microsoft stock fell 2% premarket on Nov. 13 and about 1% Monday afternoon.
  5. Final post-incident report due this week.

A software bug present in Microsoft's gateway servers since July knocked out more than 80% of the machines routing Microsoft 365 traffic at the worst point of a global outage that began on Nov. 12, the company has confirmed.

Microsoft traced the defect to a patch written for the gateway servers that direct Microsoft 365 network traffic. Under particular traffic patterns, the bug could leave servers unresponsive, according to BleepingComputer. The flaw survived months in production and every pre-release test because Microsoft's load testing ran on lab machines with weaker hardware and fewer processor cores than the servers actually deployed — hardware on which, the company said, the problem worsens on newer machines.

What set the bug off?

The chain of events started on the evening of Nov. 11. At roughly 8 p.m. ET, Teams traffic was automatically rerouted onto the gateway servers that also handle Outlook, according to Bloomberg. That surge, combined with server capacity lost to a separate problem days earlier, triggered the latent defect and drained the servers of memory. Nearly 14% of the gateway fleet's capacity was already offline, according to BleepingComputer.

Charles Lamanna, Microsoft's executive vice president of Copilot, Agents, and Platform, said the reduced capacity alone was not the cause. "Had no impact on its own," he said, according to ABC News.

The failure then compounded. As servers went down, Microsoft's systems shifted users onto remaining machines, which buckled under the extra load. Microsoft said the defect spread across the server fleet within two hours, and at the peak more than 80% of the servers had stopped responding. Safeguards designed to keep a failure in one region from crossing into others did not hold.

How long did it last?

The incident began at 11:20 a.m. ET on Nov. 12. By 11:10 a.m. ET on Nov. 13 — nearly 24 hours later — Microsoft said customers were no longer actively affected.

Recovery ran slower than in a normal incident, Lamanna said, because servers coming back online were immediately swamped by the traffic that had piled up while they were down, according to ABC News. Rolling back recent updates brought no relief: the flaw had been sitting in the code since July, and no one had noticed it.

"We are deeply sorry for the impact to all our customers and recognize the impact that this has on their normal business & operations," Lamanna wrote in a post on X, according to ABC News.

Who was hit?

The outage drew reports from 74 countries and territories on Downdetector. It disrupted the U.S. House of Representatives. Singapore's Government Technology Agency said government emails were down, and Japanese government employees resorted to printing paper copies of documents.

Investors registered the damage. Microsoft stock slid 2% before the open on Nov. 13 and was down about 1% in Monday afternoon trading.

What has Microsoft changed?

The company has since expanded gateway capacity, upgraded monitoring and introduced protections designed to ensure that rerouting traffic cannot set off a similar chain-reaction failure, according to BleepingComputer.

A final post-incident report is expected this week.

via qz.com (Source)

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Staff writer covering companies, deals and strategy in corporate strategy and chief-executive leadership at CEO Frontier.

514 articles

Filed under

Adjacent sheets

« Previous article

End of sheetN-51 / M9 · November 20, 2026