Microsoft’s 18-hour outage across 74 countries disrupted services like Outlook and Teams, forcing many to use paper and faxes. On Friday, Microsoft said the fault came from a network upgrade, not AI or hackers. The system fell over as traffic returned, showing how easily our biggest systems break down.
Microsoft's 18-hour outage across 74 countries disrupted services including Outlook, Teams, OneDrive and Copilot, forcing hospitals, police forces, airlines and businesses to fall back on paper, phone calls and fax machines.
Last week, Outlook, Teams, OneDrive and Copilot went down in 74 countries. It was close to 18 hrs before Microsoft could say things were mostly back to normal. The office jokes came fast. Microsoft 365 became 'Microsoft 364', and someone asked whether we should start sending faxes.
Then, the less funny stories came in. Some British hospitals cancelled routine appointments and went back to paper. Police forces asked people to phone in crimes because online forms had stopped working. Japanese ministry officials printed documents to keep things moving.
Closer home, airlines warned passengers that their email support was affected, and a telecom operator's call centres were flooded. In a few markets, traders confirmed deals by phone and fax.
On Friday, Microsoft said sorry. The fault, it said, came from an earlier network capacity upgrade, and only showed itself after a later network change. It was careful to add that this had nothing to do with AI or hackers. That distinction was not as comforting as it may have sounded.
On Saturday, in an unrelated incident, a network fault grounded every departing flight in the US for more than 4 hrs. Again, no attacker. Two failures in two days, with no connection between them, point the same way: these days, our biggest systems are more likely to break from the inside, during routine work, than from any outside attack. The cushion that used to absorb small mistakes has been quietly trimmed away.
You could see that in how long the recovery took. When some servers went down, their traffic piled onto the rest, and those went down too. Microsoft admitted that some servers it brought back fell over again as soon as traffic returned. Its first fixes didn't work. In the end, engineers had to bring the system back slowly, almost one piece at a time, checking each step before moving on.
A few engineers outside the company think a flaw had been sitting in the code for years, and the recent change finally woke it up. Microsoft's own report should tell us whether they're right. Whatever the root cause, getting a system like this back on its feet depends on people. Specifically, it depends on people who know how an old, sprawling system actually fits together, often in ways nobody ever wrote down. You can't hire that knowledge in a hurry, or generate it.
Which is why this isn't just a story about a network team having a bad night, but one about choices made much higher up.
Microsoft's leadership hasn't been shy about where its priorities lie. Satya Nadella has said AI already writes a meaningful chunk of the code in some of the company's projects. Microsoft expects to spend around $175 bn this year, most of it on AI infra. Earlier this year, it offered buyouts to about 9,000 US staff, then cut 4,800 jobs in July. Barely 2 wks before the outage, it offered voluntary exits to roughly 60,000 people, more than a quarter of everyone who works there.
To be fair, nobody can say these decisions caused last week's mess. Most people offered the latest buyouts haven't decided yet; they have until December. We also haven't seen Microsoft's full account. But anyone who has worked in a large tech company knows how this goes. Nobody gets promoted for keeping old systems running quietly. Right now, every incentive points towards shipping the next AI feature.
Effectively, this was an extraordinary failure for Microsoft's leadership. This is a company asking the world to trust it with the future of work, and for most of a day, it couldn't deliver email.
When Microsoft's report comes out, it will probably blame a configuration change and a few lines of bad code. It should also answer the question that really matters: who inside Microsoft still understands how this thing works? And is the company doing enough to keep them?
Will it lose customers over this? Almost certainly not. Moving a big organisation off Outlook and Teams takes a year, a lot of retraining, and an eye-watering consulting bill. No CIO is going to sign up for that because of one bad Thursday-Friday. Microsoft knows this. It's the reason a company can cut staff, raise prices and have a day like this, all in the same year, and still keep almost every renewal. When customers can't really walk away, nothing in the market forces anyone to take reliability seriously.
Someone else will have to. A small number of cloud and software companies now hold up hospitals, courts, parliaments, airlines and stock exchanges. They deserve the same scrutiny as banks did post-2008 post-Global Financial Crisis. Before the next big outage, not after it.
India should be pushing hard on this. Its ministries, banks, IT firms and airlines run on the same handful of providers. RBI already expects banks to plan for a critical vendor going down, and that thinking needs to spread well beyond banking. Large Indian buyers can also do more themselves. Write recovery drills, independent audits and real penalties into contracts. Stop treating an uptime percentage as a guarantee, because it means very little on the day everything stops.
India supplies many of the engineers who sit on these 3 a.m. incident calls, fixing systems someone else designed years ago. If anyone understands how thin the bench has become, it's us.