For more than twenty hours this week, one of the largest email systems on the planet would not answer. Microsoft 365 went down Thursday around noon Eastern, taking mail, chat, shared files, scheduling, and a long list of connected tools with it. The outage tracker Downdetector logged about 130,000 reports at the peak, and within roughly an hour reports were arriving from 74 countries and territories. Banks, courts, rail operators, and governments reported disruption, and several British police forces pointed residents toward phone lines while it ran.
No one broke in. A Microsoft executive said Friday there was no sign of malicious or cyber activity, which is what makes the story worth a business owner's time. It was the boring kind of failure, and it unfolded in public: the provider did not know what was breaking for hours, and its explanation kept changing.
Microsoft's own words, hour by hour
The first word, Thursday afternoon: engineers were investigating infrastructure experiencing "elevated load across multiple regions," restarting and repairing what they could, and "reviewing an unexpected increase in a specific type of service traffic." First theory: too much traffic, from somewhere.
The traffic theory did not hold. Later Thursday came a correction: "We've not identified any recent change correlated with the start of impact." The investigation pointed instead at servers failing automated health checks, tracing back to "a specific component that's responsible for processing server health." The status page admitted the hard part: "our actions taken thus far have not provided the anticipated relief." The new finding: "a core process on the affected servers is becoming overwhelmed, which causes the servers to stop responding and contributes to the capacity loss."
The fix evolved in public too. Microsoft paused all deployments and configuration changes so nothing under way added fuel. By late afternoon the engineers had "validated an alternative service health check method," began applying it "more broadly," and stood up "automated recovery systems to continually detect and restart unresponsive servers." To test the approach safely, they were "fully isolating a portion of our service infrastructure, restarting the affected components, and gradually reintroducing traffic." By Thursday night the diagnosis had a name: the process that establishes secure connections, the handshake every session begins with, was the one being overwhelmed.
Friday's updates shifted to region-by-region recovery. By mid-morning the company reported email queues "significantly reduced and approaching normal levels," while still warning that "service health remains variable and capacity has not yet fully returned to baseline." Mail that could not send keeps trying to deliver for up to 48 hours, so late messages arrive rather than vanish. The root cause, as of Friday morning, remained unpublished.
Everyone else guessed in real time
Customers learned the news through journalists and rumor: Engadget reported its staff could not even load Microsoft's own service status page Thursday afternoon. Posts claimed other major cloud providers were down too; those providers said, correctly, they were operating normally. Reuters reported Australia's cyber security minister comparing the incident to the 2024 bad-update outage that grounded airlines and banks. Nobody guessed in bad faith; in the hours when a business most needed answers, even the experts worked from incomplete information.
Your data survived. Your access did not.
This outage required no data recovery effort. The provider kept the mail safe the entire time. What it could not keep was the day.
For twenty hours, the thing a business ran on was access. The invoice sat in a queue nobody could open. The quote could not go out. An electric truck maker pushed its quarterly earnings release from the outage day to November 16, citing the disruption, per CNBC. When a provider fails this way, the contract typically compensates with service credits measured against a monthly fee. Those credits do not refund a lost day.
The plan worth writing today
None of this is an argument against the cloud; it remains the right choice for most businesses. But a provider going dark is a planning problem, and the answer is written in advance or not at all. The whole plan fits on one page, and it has four moves:
Map what stops the day. Walk a normal Tuesday and mark every system that stops revenue when it goes quiet: email, the calendar your bookings live in, the shared file space your crew works from, the phone system if it shares the platform. This week's outage took several of those down at once, because they share a provider.
Pick a channel the outage cannot touch. One airline said it had temporarily rerouted its email so customer messages kept flowing while the outage ran. A phone tree, a group text, or a printed contact sheet runs on the cellular network, which had a perfectly fine week. Organizations whose inbox is the mission go further: a second mailbox system from a different provider, on a separate subdomain, addresses tested in advance, so the team keeps corresponding while mail aimed at the main domain waits at the senders. The test for any fallback is simple: it must not depend on the thing that just failed, and the team has to know it exists before the morning it is needed.
Ask the failure question before you renew. Provider outages are public history, and this decade has delivered several multi-continent failures caused by ordinary operations rather than attackers. Ask whoever runs your platform how it failed the last time, how the news reached users, and what recovery took. Those answers are more useful than the uptime percentage on the sales page.
Write it where the outage cannot reach it. The plan, the phone numbers, the customer script, the list of who decides what: on paper, in a drawer more than one person can find. If the only copy lives inside the system that just went dark, the plan went dark with it. The paper also serves as the record of what happened, ready to be typed back in once systems return.
The Bottom Line
Microsoft needed twenty hours to name its own problem, and its explanation changed with every update. Plan for the same fog on your side: the provider unsure, the status page slow, the rumor fast. The difference between an inconvenient day and a damaging week is rarely the technology; it is whether the fallback already existed, and the inexpensive time to write that fallback down is while everything is healthy. If nobody on your team could say where the paper copy lives without looking it up, that is the conversation worth having with someone who does this for a living.
Story reporting: Microsoft status updates and the Microsoft 365 Status account on X, November 12 and 13, 2026; CNBC, November 12 and 13, 2026; Engadget, November 12, 2026; Reuters, November 13, 2026.