ZURICH. A worldwide failure in Microsoft’s cloud services disrupted Outlook, Teams, Copilot and other Microsoft 365 tools from Thursday evening, and it was still not fully over by Friday lunchtime. The company said individual servers kept falling even as others were stabilised, and that it had found no evidence of a cyberattack. It blamed infrastructure and capacity. The outage reached businesses, public offices and private users in at least 74 countries. Switzerland felt it in the places that run on those tools, not in the services that do not.

Office work at Swiss Federal Railways and at Swiss Post was hit. Trains and trams ran. Letters and parcels were still sorted and delivered. Migros and Coop said their core shop business was unaffected. SWISS and SRF both reported problems. Richemont, the luxury group, postponed publication of its half-year results because it could not complete the release while the tools were down. That is a useful measure of the outage: serious enough to stop a listed company speaking to the market, and not serious enough to stop a train.

What failed, in ordinary language

The trains ran. The inboxes did not. Microsoft says this was capacity, not an attack.

Email, chat, calls, calendars and some connected apps became unreliable or unavailable. Interfaces to outside providers failed at times as well. For a newsroom, a hospital administration desk or a logistics office, that is the working day. For a passenger, it was invisible, which is why SBB was at pains to separate the office outage from the timetable. The same split applied at the post office: the counters and the vans are not the same system as the staff inbox.

Microsoft’s assurance that this was not an intrusion matters, and it should be held as the company’s statement rather than as a finished forensic report. Capacity failures and attacks can look alike from a desk in Zurich: the mail will not send, and nobody can say when it will. The difference for a security team is whether data left the building. Microsoft says it has no evidence that they did. Firms that treat the outage as a breach, and start resetting every password in a panic, are responding to a different incident from the one described.

What to do with the backlog

Assume that messages sent on Thursday evening and Friday morning may have failed, rather than assuming they arrived late. If a payment instruction, a medical referral or a legal deadline depended on an email in that window, confirm it by a second channel once the service is stable. Do not confirm it by clicking a link in a surprise message that offers to “restore your mailbox”. Outages are when phishing copies the vendor’s tone. Microsoft will not ask you for a password by text.

Richemont’s delay is a reminder for any firm with a scheduled announcement. If the numbers cannot be checked because the files are in a cloud that is down, the correct market behaviour is to say so and wait, which is what the group did. Guessing the figures to keep a slot is how a technical fault becomes a disclosure fault. For everyone else, Friday’s lesson is simpler. The country’s physical systems kept their promises. The American software layer many of those systems now sit on did not, for the better part of a day, and Microsoft’s own account is that the cause was its capacity, not a break-in.