One vendor, half the planet offline: Microsoft’s global outage
A Microsoft 365 failure that began on November 12 knocked out Outlook, Teams and Copilot in 74 countries while the company’s own fixes kept failing, and restored servers fell silent again under traffic.
What happens when the world’s email runs through one company became visible on Thursday, November 12, 2026, as a Microsoft 365 failure spread from the Americas across the globe and stretched into the Asia-Pacific business day. Outlook, Teams and Copilot went down together, and the company’s own repair attempts kept failing in public.
Microsoft attributed the disruption to what it called elevated load across parts of its infrastructure in multiple regions, a description of the symptom rather than a cause. In a series of updates posted on X, the company admitted that fixes it had applied were, in its words, “not taking hold as expected,” and said it was looking for alternate workstreams to add capacity. At one point it reported only limited recovery as it rerouted traffic away from damaged servers, warning that customers would return gradually as their clients reconnected.
The scale was unmistakable even before the cause was known. Outage tracker Downdetector recorded the failure in 74 countries and territories within about an hour of it going global, according to AFP’s account of the tracker data, while reports of Outlook problems peaked at nearly 130,000 shortly after 1 p.m. Eastern Time, with tens of thousands more for Microsoft 365 itself, as Channel News Asia (CNA) reported from the Reuters wire. What Microsoft first described as an investigation into reports in the Americas was soon widened: the company confirmed the fault could affect Microsoft 365 connectivity for users in any region.
Behind the bland status language sits a more troubling admission. Microsoft said a process responsible for establishing secure connections on the affected servers had become overwhelmed, leaving servers unable to process requests, and that it was examining recent configuration changes as possible contributors, as the Australian Broadcasting Corporation (ABC) reported on November 13, 2026. Restored servers behaved worst of all: brought back into service, they fell silent again once traffic reached them. The company paused deployments, throttled traffic and began restarting its fleet, but each recovery proved partial.
Governments felt the outage as directly as companies did. In Australia, the director-general of the Australian Security Intelligence Organisation, Mike Burgess, and the Australian Federal Police commissioner, Krissy Barrett, confirmed their organizations were affected, and Home Affairs Minister Tony Burke said the Australian government had activated its cyber agreement with Microsoft to secure priority restoration, CNA reported. Australia’s national cyber security coordinator opened an inquiry, more than 4,400 Outlook failure reports were logged in the country by mid-morning, and some users found themselves locked out of unrelated services because second-factor codes never arrived, the ABC reported. New Zealand’s foreign ministry told posts worldwide, including every embassy and consulate, that its ability to receive email was affected.
Microsoft said there was no indication of a cyberattack and that it knew of no malicious activity behind the failure. The Australian government also said there was no indication of a cyberattack. That denial settles only the narrowest question. The root cause remains unnamed, and every official phrase on offer describes load, traffic and overwhelmed connection handlers without explaining what tipped them over.
A single point of failure with a global address book
The structural lesson is clear: a growing reliance on a handful of cloud providers means one vendor’s bad day becomes dozens of countries’ standstill. Intelligence agencies, federal police, foreign ministries and ordinary office workers lost the same tools in the same hours because they rent them from the same landlord. Concentration on this scale converts an internal configuration problem into a public event, and turns the vendor’s status page into a government communiqué.
The outage also exposed how much of daily authentication now leans on the same stack. When the mail system that delivers login codes stops delivering mail, the failure cascades past email into every service guarded by those codes.
Microsoft said the restored servers fell silent again once traffic reached them and that its fixes were not taking hold as expected.
The open question the episode leaves is who carries the cost of concentration. Microsoft absorbs the reputational damage and sells the restoration, while governments activate bilateral agreements to jump the repair queue and everyone else waits for gradual recovery. A market that rewards scale keeps producing single points of failure, and each outage rehearses the same answer: the provider routes traffic, the customers absorb the standstill, and nobody prices the dependence until the next one begins.