What Microsoft said first: part of its infrastructure was “experiencing elevated load.”
What fixed it: a different way of checking whether its own servers were healthy.
Microsoft Outlook went dark for a large part of the Americas on Thursday afternoon, with about 130,000 Downdetector reports landing shortly after 1 p.m. ET. Email would not send, inboxes would not load, and Teams and Copilot stumbled alongside it.
The coverage has mostly repeated the word “load.” Microsoft’s own updates tell a stranger story, and it is worth reading them in order.
The Paper Trail, In Order
We lined up Microsoft’s public statements, as reported through the afternoon, against the outage numbers. Here is the sequence:
- Morning: Microsoft says Teams problems in North America have been “successfully restored,” NBC News reported. Some users still cannot connect.
- Just after 1 p.m. ET: Outlook reports spike toward 130,000. Microsoft confirms users are “experiencing issues accessing Exchange Online mailboxes through multiple connection methods in the Americas.”
- Early afternoon: The company blames “elevated load” and says it will start “redistributing” traffic.
- 2:46 p.m. ET: Massachusetts texts state workers that “The Commonwealth’s email and other Microsoft services are currently unavailable,” GBH reported.
- Just before 3 p.m. ET: Reports fall to about 77,000, TechRepublic noted.
- Around 4 p.m. ET: Microsoft says it has “validated an alternative service health check method” that brought some servers back, and is “applying this change more broadly,” CNBC reported.
Admins can follow the incident in the Microsoft 365 admin center under EX1550543.
Load Was the Symptom
Read steps three and six together. Microsoft later said the load came from “a number of servers being unexpectedly unavailable,” with servers “failing automated health checks.” The survivors buckled under the extra traffic.
Then the fix was not more capacity. It was a new health check.
That points somewhere uncomfortable. Microsoft has not said its monitoring pulled healthy machines out of rotation, and we are not claiming it did. But when swapping the test brings servers back, the test was at least part of the problem. A cloud that guards itself with automated checks can also knock itself over with them.
Who Pays for a Single Inbox
Massachusetts is the detail that should stick. A state government told its workforce, by text message, that its email no longer existed for the afternoon. Schools, hospitals and city halls across the region run on the same Exchange Online back end.
We think that concentration is the real story. Public agencies moved to Microsoft 365 because it was cheaper and easier than running their own mail. The trade was resilience, and on days like this the bill comes due all at once, for everyone, with no local fallback.
Microsoft has said it is building automated recovery to restart unresponsive servers. That helps Microsoft. It does nothing for an agency with no second way to send an email.
What Users Can Expect
Mail sent during the outage is not necessarily lost. Exchange Online keeps retrying delivery for up to 24 hours before bouncing a message back to the sender, so a backlog may arrive in a rush.
Amazon Web Services said it was “operating normally” and that reports of an AWS outage were inaccurate. Google users also logged complaints on Downdetector, though no wider Google failure was confirmed.
As of Thursday evening, Microsoft had not declared full restoration or published a root cause. When it does, the question to ask is simple: did the servers fail, or did the checks say they had?