Microsoft's cloud broke down at 13:26 ET on Thursday. Mail, chat and its Copilot assistant all went dark at once1.
The company's own fix fell short. Twice. By Friday morning, Microsoft's notes still described a partial recovery16.
Anthropic then logged trouble of its own. One problem traced straight back to Microsoft. A second, separate one stayed unexplained23.
Microsoft's Fix Fell Short, Twice
Microsoft pointed to its own servers as the cause, separate from any rise in traffic. "We determined the elevated load isn't driven by an increase in traffic, but by a number of servers being unexpectedly unavailable," the company said1.
Engineers paused other deployments. They spread traffic across regions and restarted servers that had dropped out1. Some service returned by 18:57 ET Thursday. Full recovery waited for hours more.
At 22:13 ET Thursday, Microsoft's own status account posted an update. "We've monitored the status of the changes applied to improve serviceability and confirmed they're not taking as expected," it wrote6.
Microsoft 365 Status@MSFT365StatusWe’ve monitored the status of the changes applied to improve serviceability and confirmed they're not taking as expected. We’re identifying alternate workstreams that can be applied to increase capacity in the face of sustained traffic management imbalances. More info under…
A Hacker News thread tracked the stretch in its own title: "Microsoft Outlook/365 still down after 12 hours"5. Outlook, Teams, SharePoint, Defender and Azure all carried pieces of the failure1.
Microsoft had already faced a smaller Teams incident that same morning, from 09:45 to 10:53 ET. The company linked that one to resource contention, a separate cause from the afternoon's bigger failure1.
5 Posted Updates Span Two Companies in Under 24 Hours
Key posted updates from Microsoft and Anthropic, Nov. 12 to 13, 2026, Eastern time
Each pin is one posted update. The blue pins are Microsoft's outage, and the magenta pins are Claude's separate sign-in trouble the next morning.
Microsoft and Anthropic post times in different formats; this piece converts all five to Eastern time from the companies' own status pages and posts.
Sources: Bleeping Computer [1]; Anthropic Status [2]; Startup Fortune [3]; Microsoft 365 Status on X [6]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Date | Event |
|---|---|
| Nov. 12, 2026 | Outage begins. Began at 13:26 ET, by Microsoft's own account |
| Nov. 12, 2026 | Mail connector down. Anthropic traced its mail and calendar connector's trouble to Microsoft at 17:23 ET |
| Nov. 12, 2026 | Shortfall update posted. Microsoft posted at 22:13 ET that its serviceability change fell short of plan |
| Nov. 13, 2026 | New sign-in bug. Anthropic opened this one at 04:58 ET, apart from the connector trouble |
| Nov. 13, 2026 | Sign-in fix ships. A fix went live at 06:05 ET, and Anthropic watched for it to hold |
Claude Had Its Own Bad Morning, Twice Over
Claude's trouble started as a side effect. Anthropic runs a Microsoft 365 connector inside Claude, a bridge of code that lets Claude read a user's mail and calendar. That bridge failed Thursday evening, and Anthropic confirmed the cause by 17:23 ET: Microsoft's own outage3.
Friday morning brought a second problem. This one stood apart from the connector, at least on paper. Anthropic opened it at 04:58 ET. "We are investigating an issue with users authenticating to Claude.ai and Claude Code," the company wrote2.
A fix went out by 06:05 ET, and Anthropic moved to watching for it to hold2. Its notes on that second incident left Microsoft unnamed. The two Claude problems shared a morning. Their causes diverged, at least in what Anthropic has published.
The Numbers Behind the Noise
Downdetector logged more than 130,000 reports for Outlook alone during Microsoft's outage3. Claude's own sign-in trouble drew 2,906 reports, against a normal day's count of two4.
Outlook's Outage Drew 45 Times Claude's Downdetector Reports
Downdetector problem reports logged during each company's own outage, Nov. 12 to 13, 2026
Each column is one outage's Downdetector report count. Outlook's column covers a much larger base of daily users than Claude's.
Outlook and Claude have very different numbers of daily users, so the gap reflects audience size as well as the severity of each outage.
Sources: Startup Fortune [3]; Ascendants [4]. Chart by Ryan Elliott Dennis.
The numbers behind this chart
| Item | Value |
|---|---|
| Outlook | 130,000 |
| Claude | 2,906 |
Both numbers measure the same thing: people hitting a button that stopped working. Outlook has a much bigger button. Microsoft's cloud suite reaches a far larger crowd of daily workers than Claude's newer, smaller base of users.
Trading Desks Reach for the Phone
Saurabh Verma, a vice president at the consultancy Frost & Sullivan, watched the trouble spread past email alone. "The continuing outage of Microsoft's collaboration tools, especially emails and scheduling, will not only impact communications, but also have a cascading impact on decision-making, revenue and service delivery," he said7.
Vijay Valecha felt that spread from a trading floor. He runs investments at Century Financial, a brokerage based in the United Arab Emirates.
"Institutional execution desks lost their primary email channels during a turbulent week of trading and were therefore forced to revert to phone calls and WhatsApp to confirm settlements," Valecha said7.
That is the plain cost of a cloud outage. Files stay safe. Teams lose the one channel everybody agreed to use, right when the week needed it most.
Does One Bad Night Mean You Need a Backup Plan?
Ahmed AlZarouni, a technology advisor in the UAE, draws the lesson in blunt terms. "We cannot afford to outsource our entire operational capability and resilience to a single provider ... business continuity must be treated as both a technical and a human failover," he told The National7.
Andrew Baker takes the opposite lesson from the same kind of night. He runs technology for Capitec Bank, a South African lender, and spends his blog arguing against the fix AlZarouni wants.
"Resilience is achieved through isolation and simplicity. Multicloud offers neither. If you cannot survive a failure in one cloud, adding a second will not save you," Baker wrote8.
AlZarouni's case holds if Friday's overlap repeats. Two providers failing in the same stretch, by his reading, means a third failure is only a matter of time. Baker's case holds regardless. His point is that a badly built pair of providers loses to one provider built well, whatever else breaks around it.
Both can be right together. A company can need a real fallback for the hours a cloud tool goes dark. That second cloud, barely known to the team running it, just adds cost.
What to do
Write down your fallback for mail and chat today, before the next outage arrives. Valecha's desks reached for WhatsApp and phone trees only after Outlook went dark7. A plan made in a calm week beats one improvised in a loud one.
Ask your own AI vendor which outside services its tools quietly depend on. Claude's mail and calendar connector failed because Microsoft's cloud failed under it, a weak link Anthropic inherited3. The same chain likely runs through tools your own team uses daily.
By the numbers
- Microsoft's outage began at 13:26 ET Thursday, by the company's own account1.
- Outlook alone drew more than 130,000 Downdetector reports during the outage3.
- Claude's sign-in failure drew 2,906 Downdetector reports, against a normal count of two4.
- Limited recovery arrived at 18:57 ET Thursday, hours short of a real fix1.
- Anthropic needed from 04:58 ET to 06:05 ET to patch Claude's separate sign-in failure2.
- 5 Microsoft services carried pieces of the failure: Outlook, Teams, SharePoint, Defender and Azure1.
What to watch
Microsoft typically closes a major outage with a public account of the server change that triggered it. That account is still pending1.
Anthropic's own notes keep Friday's sign-in failure apart from Thursday's connector trouble. A future update linking the two would make AlZarouni's argument stronger than the record shows today2.
Sources
- Lawrence Abrams, "Massive Microsoft 365 outage disrupts Outlook, Teams, Copilot worldwide," Bleeping Computer, Nov. 12, 2026, https://www.bleepingcomputer.com/news/microsoft/massive-microsoft-365-outage-disrupts-outlook-teams-copilot-worldwide/
- "Elevated errors on claude.ai, Claude Code, and platform.claude.com," Anthropic Status, Nov. 13, 2026, https://status.claude.com/incidents/nn11wk485zfn
- Hiro DXB, "Claude Suffers Login Outage Across Claude.ai and Claude Code While Microsoft 365 Connector Stays Down," Startup Fortune, Nov. 13, 2026, https://startupfortune.com/claude-suffers-login-outage-across-claudeai-and-claude-code-while-microsoft-365-connector-stays-down/
- Heer Shah, "Claude AI Down Today? 2,906 Outage Reports On Nov 13," Ascendants, Nov. 13, 2026, https://ascendants.in/industry_events/claude-ai-down-november-13-2026-outage/
- "Microsoft Outlook/365 still down after 12 hours," Hacker News, Nov. 13, 2026, https://news.ycombinator.com/item?id=50478609
- Microsoft 365 Status, post on X, Nov. 13, 2026, https://x.com/MSFT365Status/status/2121079061774491735
- Aarti Nagraj and Alvin R Cabral, "Microsoft outage to have 'cascading impact' on businesses," The National, Nov. 13, 2026, https://www.thenationalnews.com/future/technology/2026/11/13/microsoft-outage-to-have-cascading-impact-on-businesses/
- Andrew Baker, "Why Multicloud Is Not a Cloud Resilience Strategy," andrewbaker.ninja, Mar. 27, 2026, https://andrewbaker.ninja/2026/03/27/why-multicloud-is-not-a-cloud-resilience-strategy/
