Microsoft 365 Outage Runs Past 12 Hours With Insured Losses Estimated at $300M to $500M

Microsoft'sMSFT email and calendar services failed for more than 12 hours this week, and the cost landed on trading desks and government agencies that had to fall back on fax machines and paper records. The outage began Thursday and was still being repaired on Friday, a year after Amazon's cloud went dark.

Microsoft said the connectivity problems stemmed from an issue during an earlier network capacity upgrade, and that it saw no sign of malicious activity. Nearly 130K US Outlook problem reports hit Downdetector at the peak, though those are complaints rather than confirmed users.

The damage spread through the plumbing of daily business. Government agencies in Australia, Japan, and the US lost email, airlines warned customers about slow communication, and trading firms saw delays. One Taiwanese financial company canceled a bond repurchase transaction, another used fax to confirm trades, and WorkhorseWKHS pushed its results to Nov. 16. In Britain, a trial in Manchester was halted and jurors were sent home, while police asked people to call in urgent reports.

Morningstar DBRS vice president Mario De Cicco put global insured losses at $300M to $500M, below the $400M to $1.5B toll from the 2024 CrowdStrike incident.

Backup systems keep failing first

Outlook and Teams serve hundreds of millions of users, and the systems meant to catch a failure keep failing too. Microsoft said some servers were failing automated health checks, and a different health check method brought some of them back. The same pattern runs through the year. In February, an Azure West US outage lasting 20 hours and 26 minutes traced back to a transformer fault and a generator switch control failure that drained the backup batteries. Google Cloud's London region went dark for six hours and eight minutes in October after backup generators couldn't start because of a generator controller fault.

The longest completed failure of the year ran 22 hours and six minutes in Azure's West US 2 region in May, after lightning-driven voltage fluctuations overwhelmed cooling systems. AWS lost a zone in Northern Virginia that same month, and it took ~21.5 hours to recover most services after a thermal event cut power.

A year earlier, AWS failed the same way

Amazon'sAMZN cloud failed on Monday, Oct. 20, 2025, when an outage in its US-EAST-1 region took down thousands of apps and websites. It began in the early morning, and services didn't return to normal until 6 p.m. Eastern.

Amazon's own review traced the failure to a flaw in the automated system that manages DynamoDB's domain name records, which left the service's main address empty. Problems then spread to its load balancers and new server launches. Tenscope estimated that global businesses lost ~$75M per hour while major sites were offline. Insurance covers less of that than most companies expect, since many cyber policies don't trigger unless an outage lasts eight hours or more, McGill and Partners' Ryan Griffin said.

Burgess said users can't pinpoint the cause because they see Snapchat or Roblox rather than Amazon. eToro analyst Josh Gilbert said repeat failures give every IT department a reason to hesitate before loading more work onto one provider, though switching costs argue the other way.

Cloudflare followed a month later, on Nov. 18, 2025, with its own global outage, and this year's outage tracker lists 11 major incidents across AWS, Microsoft, Google Cloud, and Cloudflare.

Microsoft said it saw no sign of an attack, and Amazon traced its outage to its own network monitoring. In both cases, the part built to catch failures was part of the failure. Companies that run their email, chat, and AI tools through one provider have little to fall back on when that provider's safety net gives way.