For more than 24 hours, one of the most relied-upon systems on the planet simply stopped working.
Continue reading to the part where Microsoft plans on splitting compute work loads amongst devices we already own. The patents have been filed.
The Microsoft 365 outage that began Thursday morning knocked out email, calendars, Teams and Microsoft’s Copilot AI for businesses, governments, courts and hospitals in 74 countries and territories.
By the time Microsoft said Friday that it had fixed the underlying problem, economists were putting the bill in the billions of dollars and the number of people who lost real work in the tens of millions.
Microsoft says its own network upgrade caused it.
The Damage by the Numbers
Up to 51 million people suffered genuine economic losses on Friday alone, and as many as 26 million on Thursday, according to Patrick Anderson, president of Michigan-based Anderson Economic Group, who spoke to CNN. Anderson said most of Microsoft’s roughly 430 million users would experience the outage as an annoyance they could work around, but tens of millions could not.
His firm estimated lost productivity at about $7 billion and said the figure was likely to climb.
Cyber-risk firm Kovrr put economic losses higher, at about $16 billion on a lost-productivity basis, according to Shalom Bublil, the company’s co-founder and chief product officer. Kovrr estimated insured losses at around $700 million. Mario De Cicco of credit rating agency Morningstar DBRS put insured losses at $300 million to $500 million.
The outage reached 74 countries and territories within about an hour, according to outage-tracking site Downdetector, spanning North and South America, Europe, Asia, Oceania and parts of Africa. At its peak, Downdetector logged 19,650 problem reports per minute across Microsoft’s email, Teams, OneDrive, Copilot and Microsoft 365 services.
Those reports capture only people who took the time to file one. The true number affected is exponentially larger.
About 3.9 million businesses worldwide use the Microsoft suite, many with no reliable backup system, said Ashwin Pal, a risk advisory partner at RSM Australia. He told ABC News Australia that losses for professional firms such as law and accounting offices would run into the hundreds of thousands of dollars per hour. “It’s billions of dollars,” Pal said.
Microsoft shares were down 0.4% in early trading Friday, Reuters reported.
Governments, Courts and Hospitals Go Dark
The failure reached the highest levels of government on several continents.
In Japan, Deputy Chief Cabinet Secretary Masanao Ozaki said the outage hit the national government’s IT infrastructure. The Philippines’ Supreme Court said its operations, including email, were affected. The U.S. House of Representatives said its users were affected, and Britain’s Parliament said it was working with Microsoft and using other ways to communicate.
In Britain, the outage hit the email system of the National Health Service, and the National Pharmacy Association warned patients could face delays receiving prescriptions. In Toronto, Unity Health Toronto declared a “Code Grey,” the alert hospitals use for a failure of infrastructure, though it said patient care was not affected.
In Southern California, Los Angeles County said the outage was blocking county departments from email and key communications platforms. The Sheriff’s Department, the Los Angeles Police Department, the Superior Court, the Medical Examiner’s Office, Culver City, the Department of Water and Power and Metro all reported email problems Thursday. The City of South Pasadena posted a service alert saying incoming and outgoing email could be delayed.
Markets and Business Scramble
As the outage rolled into the European and Asian workday Friday, the business damage spread.
Euronext, which runs stock exchanges in Paris, Amsterdam, Brussels, Lisbon, Milan, Oslo and Athens, told listed companies to pick up the phone. At least one bond repurchase in Europe was halted, and some organizations went back to fax machines, CBC News reported. British insurer Aviva sent out alternate phone numbers for its earnings call and used Bluetooth to move recordings because email was down.
Trading firms reported delays, some companies could not publish earnings and airlines warned customers of communication lags, Reuters reported. Electric truck maker Workhorse Group postponed its quarterly earnings release until Monday.
Small businesses took some of the hardest hits. Tracey Tucker, owner of Regal Shower Screens in Queensland, Australia, told ABC News Australia that one lost day would cost her company at least 30,000 Australian dollars.
The outage even broke rival AI products. OpenAI and Anthropic both reported that the tools connecting their AI assistants to Microsoft 365 stopped working.
Organizations that run Microsoft’s email software on their own servers, rather than in Microsoft’s cloud, were largely spared, according to Reuters.
No Computers Crashed. That Was the Problem.
Unlike the 2024 CrowdStrike disaster, which crashed about 8.5 million Windows computers and grounded flights worldwide, not a single user’s computer failed this time.
Every laptop, desktop and phone worked fine. The breakdown happened inside Microsoft’s own data centers, where servers dropped out of service one after another until the remaining machines could not carry the load. Users’ devices were simply connecting to nothing.
It Should Not Have Happened. It Has Happened Before.
On Friday, Microsoft said the incident “was introduced by an issue during an earlier network capacity upgrade” that surfaced after a more recent network change.
That is the second time this year Microsoft has tied a major Microsoft 365 failure to capacity work. In January, Microsoft 365 went down across North America for nearly 10 hours, and the company blamed heavy load combined with reduced capacity during maintenance, Al Jazeera reported.
Microsoft’s account of this week’s outage also shifted. On Thursday afternoon, the company said it had not found any recent change that lined up with the start of the problem. By Friday, it said a network change had triggered it.
The outage tracking service UptimeRobot lists 15 Microsoft 365 incidents in 2026, most of them smaller. Microsoft also suffered a major Azure cloud outage in October that spilled into its productivity software.
Tech commentator Goodwins told Al Jazeera that Microsoft has been focused almost entirely on artificial intelligence, and that services like email and Teams may now be treated as a “forgotten revenue source.”
“We are deeply sorry for the disruption this incident has caused our customers and partners,” Microsoft said Friday.
Analysis: Microsoft Is Already Looking Beyond the Data Center
While Microsoft’s central cloud was failing this week, the company’s recent patent filings point toward a different model: pushing more AI work out to the devices people already own.
On Nov. 5, the U.S. Patent and Trademark Office published a Microsoft application titled “Opportunistic Micro-training on Edge Devices Using Spare GPU Utilization.” It describes splitting AI model training into small “micro-tasks” and sending them to opted-in smartphones, laptops and gaming computers when their processors sit idle. Owners could receive credits for the computing work, according to the filing. The crypto and tech news site TokenPost reported the patent was granted Nov. 10.
The same day, Microsoft published an application for running one general-purpose AI model on a user’s device and teaching it new skills through small downloadable add-ons. Patent-tracking site Patentlyze said it was the 21st Microsoft filing it has tracked since August related to on-device AI.
Earlier filings published in July and September describe splitting AI work between a device’s own chips and handing a running application back and forth between a local machine and the cloud.
The filings raise questions the documents do not answer. The training patent does not say how training data would be handled on consumers’ devices or whether owners would be paid for the electricity their machines use. A patent is not a product, and Microsoft has announced no service based on these filings, most of which were submitted in 2025.
Nothing connects the filings to this week’s outage. But taken together, they show Microsoft exploring a future in which more of its AI load runs on customers’ hardware rather than in its own data centers.
How The Rolling Outage Played Out
Microsoft opened an incident at 8:17 a.m. PST Thursday for users who could not reach their Exchange Online mailboxes. Exchange Online is the Microsoft system that runs email and calendars for most businesses, schools and public agencies using Microsoft 365. The problem spread through the day to Teams, SharePoint, Copilot, Planner and a growing list of business and security tools.
Microsoft said a core process on the affected servers became overwhelmed, causing servers to stop responding and fail the automated health checks that decide whether they can carry traffic. As servers dropped out, the rest absorbed more load. Servers Microsoft restored Thursday afternoon and evening repeatedly failed again once traffic returned.
Charles Lamanna, Microsoft’s executive vice president for Copilot, Agents and Platform, said late Thursday there was “no indication this is the result of malicious or cyber activity.” Microsoft later called it “a technical issue not related to cyber or AI / agents.”
Recovery began overnight. Around 2 a.m. PST Friday, Microsoft reported early stabilization. By 6:17 a.m. PST, it said email queues were approaching normal levels. At about 8:57 a.m. PST, the company said it had “addressed the underlying issue” and was restoring service to remaining customers and regions.
As of Friday morning, Microsoft’s service status page still listed Microsoft 365 for business customers under “service degradation.” Some users could still see intermittent access problems and late email. Microsoft said Teams calls, meetings and chats were working, but calendar scheduling and other email-linked features remained affected.
Microsoft has said it will publish a post-incident report with a full root cause.
Where Microsoft Customers Sit Now And What To Do
Check whether anything important you sent Thursday or Friday actually arrived. Microsoft extended how long it holds undelivered email to 48 hours, according to the tech news site Petri, so many delayed messages should still arrive rather than bounce. If a message bounced back to you, send it again.
Expect a burst of late email and notifications as apps catch up.
Watch for scams. Fraudsters often exploit big outages with fake “service restored” messages asking users to sign in again. Go directly to Microsoft’s status page at status.cloud.microsoft instead of clicking links.
Business owners who lost work should document what stopped working and when. Microsoft’s business service agreements allow for credits, but an administrator generally must file a claim.
