For almost a day, a considerable part of the world economy has returned to working as it did in the 90s, without email or online meetings. A crash of Microsoft 365, Microsoft's professional tools, has left programs such as Outlook or Teams inaccessible. Traffic began to recover at midday on Friday. By mid-afternoon, Microsoft still had the incident open with around twenty services marked as degraded.
Just one hour after the first falls, the Downdetector service was already registering incidents in 74 countries and territories. The origin of the failure seems to be in Exchange Online, the cloud mail and calendar engine on which Outlook works but which Teams uses to schedule meetings, and Copilot and Defender, to read messages and monitor security.
The impact has been enormous because Microsoft 365 has more than 430 million users around the world. Some 3.9 million companies work with these tools.
The most striking effect has been seen in the financial markets. In Taiwan, several managers, insurers and securities houses suffered problems in fund management. According to sources cited by Reuters, one of them even canceled a bond repurchase operation and another had to confirm operations by fax, a technology that many interns from those same firms have probably never used.
The results schedule also suffered. Workhorse, an American manufacturer of electric trucks, postponed to November 16 the presentation of accounts that had been scheduled for Thursday. Richemont, the Swiss group that owns Cartier, also briefly delayed the publication of its half-year results on Friday. Aviva activated backup systems for its analyst conferences and Euronext reported problems with mail.
The list includes other sectors. IndiGo, Akasa Air and Air India were left without customer service emails. In the United Kingdom, the national health service in some centers resorted to paper records and the London Metropolitan Police has been unable to receive some complaints online. There were also affected administrations in Australia, Japan and the United States. Even OpenAI and Anthropic reported failures in the connectors that link ChatGPT and Claude with Microsoft accounts.
Microsoft's version of the reasons has changed as the hours have progressed. First he spoke of a high load in several regions. He later specified that there was no more traffic than normal, but rather that a significant number of servers had become unavailable unexpectedly and the rest could not absorb the work.
From there, attention has focused on the automatic checks that Microsoft uses to decide if a server is able to receive traffic. The company went so far as to point out that a process responsible for establishing secure connections became saturated and left machines unresponsive. On Friday it was reverting a change in network configuration, although without confirming that it was the culprit.
What seems almost certainly ruled out is that it is an intentional attack. Charles Lamanna, executive vice president of Microsoft, ruled it out on social network X. The company has promised to publish a report with the root cause once the incident is resolved.
Microsoft shares fell 1.7% before Friday's open on Wall Street after hours of non-stop rulings.
For Microsoft, such a prolonged failure can have a high cost. If the monthly availability of a service drops below 99.9%, companies, educational centers and administrations can claim a credit of 25% of the fee, which can reach 50% or 100% in the most serious cases.
What worries analysts most is the pattern. Microsoft 365 already fell in January of this year and in October 2025, and last month a failure of Azure, the company's cloud, also affected its productivity suite. If reliability starts to seem like a recurring problem, IT departments will have reason to think twice before putting more business in the hands of a single supplier.
Other analysts cited doubt that these episodes push customers towards the competition, because replacing tools that are so integrated into day-to-day life is expensive. Organizations that kept Microsoft software on their own servers, rather than in the cloud, fared better.
Even so, the incident reopens the debate about the scarcity of alternatives for cloud services and the excessive dependence on American companies (Microsoft, Google and Amazon share the market almost completely). In Europe, Denmark has begun to test alternatives to Microsoft, in fact, for reasons of digital sovereignty. This week's fall adds a practical argument to the political one. You don't need a commercial dispute to run out of mail, just a server failure is enough.
