Microsoft 365 went down shortly after noon ET on Thursday and stayed unreliable for more than a day. Outlook, Teams and Copilot were the visible casualties, and Microsoft's own status page was offline until about 3 pm ET, so for the first hours customers had nothing official to read.

By Friday afternoon the company was still hedging. "As infrastructure stabilizes, we're still draining queued messages and rebalancing traffic," Microsoft posted on X. "We do not have an ETA for full recovery yet. Service access can recover before mail flow fully normalizes." It also extended the window in which queued mail would be held to 48 hours.

The damage had already spread well past unanswered email.

Courts, jails and police stations

In Maine, the judicial branch warned that requests for protection orders sent by email after 12:30 pm Thursday might never have reached a court. New York City suspended evictions scheduled for Friday, a decision the Department of Investigation said it made after consulting the Civil Court. In Manchester, a criminal trial was adjourned for the day and staff went looking for a fax machine, The Guardian reported.

In Colorado, the Arapahoe County Sheriff's Office said its jail could not process online bond payments because staff were not receiving confirmation emails, so anyone seeking an inmate's release had to pay in person. The Weld County Sheriff's Office asked residents to call or text instead of emailing.

California took the widest hit. The Los Angeles County Sheriff's Department reported disruption to email and other communication platforms, and the Los Angeles Police Department and the county district attorney's office had mail problems too. In Florida, St. Lucie County confirmed that its sheriff's office and fire district were affected.

British forces including Dorset, Cumbria, South Wales, Wiltshire and Dyfed-Powys had trouble receiving online crime reports, according to The Sun. Emergency phone lines kept working.

Hospitals fell back on paper

Unity Health Toronto, which runs several hospitals, declared a Code Grey, the alert used for infrastructure failure; it said patient care was unaffected and facilities stayed open. Toronto's Hospital for Sick Children warned that some calls and emails might not arrive.

Britain's National Health Service lost access to email, appointment information and patient data in places, with some services reverting to paper records, and the National Pharmacy Association reported prescription delays where pharmacies could not reach doctors. VIA Rail and National Rail told passengers to fetch tickets through apps or retailers rather than email.

Governments outside the Anglosphere were in the same position. Australia's intelligence organisation and federal police confirmed disruption. Singapore's Government Technology Agency reported problems with government systems while emergency services and Singpass stayed up. Japan's foreign ministry could not send or receive mail. The Philippines' Supreme Court warned of delays.

On Capitol Hill, staff lost email and calendars. "There goes a day of legislative work," one House aide told Politico. Austin, Jacksonville, Richmond and Laurel, Maryland, all pointed residents to other channels.

Microsoft says an upgrade did it

The company has ruled out an attack and ruled out its own AI agents. "This was a technical issue not related to cyber or AI/agents," a spokesperson told Wired. "The incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change."

That account differs from the early one. On Thursday afternoon Microsoft said elevated load came from "a number of servers being unexpectedly unavailable," not from more traffic. Microsoft told Australia's home affairs minister, Tony Burke, that the pattern resembled CrowdStrike in July 2024, when a bad update took down hospitals and airlines. A full post-incident report has been promised.

The bill is being counted. One think tank told CNN the outage cost businesses about $7 billion, a figure that will be argued over for months. Downdetector logged nearly 83,000 reports of Outlook problems by 1 pm ET on Thursday and was still recording around 3,000 at midday Friday, which is a measure of complaints rather than of affected seats. The University of Sussex warned students and staff about callers posing as Microsoft engineers and asking for passwords on the pretext of fixing email.

The structural point is older than this week. A 2022 survey commissioned by Google found 84 percent of government employees in the Washington area used Microsoft products as their primary suite, and a 2021 Omdia report cited an 85 percent share of the US government productivity market. Microsoft reported more than 450 million paid commercial seats in January, so a single change in one network path can reach a large share of the world's desks at once.

The company has spent the past year shrinking what those seats include while raising prices on some tiers. Consolidation is what makes the price work. It is also what made Friday's briefing in Manchester necessary.

Source: Fortune, ZDNET, Wired