Microsoft 365 outage in November 2026: what happened and what SMEs can learn from it
From the evening of Thursday 12 November 2026 until Friday afternoon, Exchange Online was disrupted worldwide. Along with Outlook, parts of Teams, SharePoint, Copilot, Intune and Defender went down. In Switzerland it mainly hit Friday morning. It was not an attack. On 15 November Microsoft explained the cause in a preliminary report: a defect from an update in July that only surfaced under particular load. This article puts the outage in context, shows how Switzerland was affected and what an SME can prepare so that the next outage is only a nuisance.
- According to Microsoft, Exchange Online was disrupted for almost 24 hours, from Thursday 17:20 to Friday 17:10 Swiss time. Reports came from 74 countries.
- Not an attack. According to Microsoft's preliminary report, a defect from an update in July took down the network gateways, triggered by more load and less spare capacity. Microsoft has announced the final report for 20 November.
- Phishing emails using the outage as bait arrived while it was still going on. During a Microsoft outage, no employee needs to click anything.
- An SME does not need a multi-cloud data centre. It needs a tested second channel, status alerts outside Microsoft and a rehearsed emergency plan.
What happened on 12 November
On the evening of Thursday 12 November 2026, users around the world reported that Outlook was no longer loading emails. Shortly afterwards Microsoft confirmed problems with Exchange Online, first in North and South America, a little later globally. Within an hour Downdetector received reports from 74 countries, at the peak almost 130,000 for Outlook alone.
Email was not the only thing affected. According to Microsoft, the disruption also hit the parts of other services that work together with Exchange Online, for example in Teams, SharePoint, OneDrive, Copilot, Defender, Purview and Intune. Most of them remained reachable, however. No outage of the Azure platform or of sign-in through Entra ID was reported.
According to Microsoft, the disruption lasted almost 24 hours, from Thursday 17:20 to Friday 17:10 Swiss time. After that, backlogs of emails still had to be cleared, and the last update came on Saturday morning. Network analyst Kentik puts the hard outage for most users at around 18 hours. Measured by duration, the outage exceeds every major cloud outage of the past two years, including AWS and Azure Front Door in October 2025.
Timeline of the outage
All times in Swiss time (CET), based on Microsoft's report. The Downdetector figures come from the press.
-
Thu, 12 Nov, 02:07
Teams moves to the Exchange gateway
Automation shifts Teams traffic onto the gateway servers that also serve Exchange Online. The load rises over the day, and individual servers already fail.
-
Thu, 17:20
Exchange Online goes down
An alert triggers the first post on the Service Health Dashboard. Outlook, Outlook on the web, ActiveSync and the APIs are affected. Around 19:00 Downdetector receives almost 130,000 reports for Outlook, from 74 countries.
-
Thu, 20:00
Microsoft stops all changes
As a precaution, Microsoft stops all ongoing and planned deployments. Shortly afterwards it is clear: the problem is not more traffic, but many failed servers whose load overwhelms the rest.
-
Overnight into Fri
Countermeasures only partly work
New health checks, restarts and redirection help briefly. Restarted servers fail again under the waiting traffic. Around 02:00 Microsoft finds the overloaded process for setting up connections.
-
Fri, 13 Nov, 08:20
The faulty code is reverted
An update restores the previous state, first on part of the servers, then group by group. Regions in business hours come first.
-
Fri, 12:35
Europe available again
Availability in Europe is healthy again, and the mail queues are clearing.
-
Fri, 17:10
Disruption over
According to Microsoft, there is no more active impact. After extended monitoring, the last update follows on Saturday morning.
Update of 16 November: Microsoft explains the cause
On 15 November Microsoft published a preliminary post-incident report. It replaces the explanations that changed several times during the outage. According to Microsoft, three things came together:
- A dormant defect from July. An update for the network gateways through which many Microsoft 365 services run contained a defect in an internal Windows process. Under certain load, processes blocked each other until processor and memory were exhausted. For more than three months the update ran worldwide without problems. Load testing did not find the defect because it ran on weaker hardware. It does not affect other Windows installations.
- More load from a normal failover. About 15 hours before the outage, Teams automatically switched to the gateway that also serves Exchange Online. Such failovers are part of normal operations.
- Less spare capacity. Almost 14 percent of the gateway servers were out of service because of an unrelated issue from the previous day. On its own, that would not have caused any disruption.
At peak time in the US the first servers failed. Automation redirected their traffic to the rest, which then failed as well. In less than two hours more than 80 percent of the servers were unresponsive. The partitions between regions that are meant to contain such a failure did not hold because of a further defect. Teams, OneDrive and SharePoint could switch to other gateways and therefore recovered earlier. Exchange Online cannot do that because of its size.
The repair took long because every restarted server was immediately overrun by waiting traffic. A defect that had been in production for months could not simply be rolled back either. The report does not mention an attack, and Microsoft had already ruled one out during the outage. Microsoft has announced the final report with all improvements for 20 November.
What the outage shows about resilience
It is striking that the very mechanisms meant to make Microsoft 365 robust made the outage bigger: the automatic failover of Teams and the redirection to healthy servers. Redundancy protects against the failure of individual parts, but not against a defect that sits everywhere at once.
This puts the outage in a series. AWS in October 2025, Azure Front Door and Cloudflare in the same autumn were also caused by changes at the provider, not by attackers. For an SME this means: the biggest availability risk is often an error at the provider, over which you have no control. So do not only plan for the failure of individual systems, but for a whole day without the service.
How Switzerland was affected
Because the outage started in the European evening, Switzerland experienced it mainly as a Friday morning without reliable email. Only around midday did Microsoft report healthy availability for Europe again, and delayed emails kept arriving into the afternoon. There were delays, but no standstill. Richemont published its half-year results 40 minutes later than planned. Swiss Post redirected urgent customer requests to the phone, while sorting and delivery ran normally. At Zurich, the German unit was at times unable to send or receive emails. Nestlé reported delays in some processes, and the factories kept running. ETH Zurich had problems with Outlook, Teams and SharePoint.
The federal administration was hardly affected. It runs its email in its own data centre and not in the public Microsoft cloud. Things were very different in the UK: there, NHS hospitals kept patient records on paper at times, and operations were cancelled.
The comparison shows what matters. Not every organisation needs its own data centre. But every organisation should know which processes depend on email and how it bridges a day without email. That is the core of emergency planning as required by the ICT minimum standard under Respond and Recover.
Phishing in the wake of the outage
More dangerous than the outage itself was the time after. Just a few hours after it began, security provider Field Effect observed fake emails for resetting the Outlook password and supposed changes to multi-factor sign-in. On 13 November a campaign followed with the subject "Outage", claiming that held messages would be deleted and leading to a fake sign-in page through a "Restore messages" button. To appear credible, it quoted the publicly known incident number. On top of that came text messages with changed bank details along the lines of "my email isn't working, please transfer to this account", and fake calls from IT.
Making matters worse, Defender for Office 365, and with it email filtering, was disrupted along with Exchange. Three simple rules apply for awareness in the team:
- During a Microsoft outage, nobody needs to do anything. Microsoft informs through the status page and the admin centre, not by email with a button to restore something.
- An incident number proves nothing. It is public, and any attacker can copy it.
- Change bank details, passwords and MFA only after a call-back, to a known number, never to the one in the message.
To practise these rules, use Phishing simulations with an outage theme. Employees report suspicious emails directly to Phishing analysis. After the outage, admins should also check the sign-in logs for new locations, new MFA methods and new forwarding rules.
Manage Microsoft as a supplier, not as a force of nature
For most SMEs, Microsoft 365 is email, telephony, file storage, device management and email protection in one. That makes Microsoft the most important IT supplier, and it belongs in the supplier register like any other. In the EU, the financial supervisors have classified Microsoft as a critical ICT third-party provider under DORA. Customers from the EU are therefore increasingly asking Swiss suppliers too how they handle such dependencies.
A look at the contract is sobering. Microsoft promises 99.9 percent availability per month, which allows around 43 minutes of downtime. If the figure is not met, there is a credit on the licence fee on request. Lost revenue, overtime and missed deadlines are not covered. Some cyber insurance policies also only pay for business interruption after an attack, not when a service provider fails. Ask your broker about this specifically.
In CyberTell's Third-party management you record Microsoft with likelihood, impact and an immediate measure if it happens. If you add Microsoft's status page, CyberTell reads its state regularly and reports the start and end of an outage to your admins by email and push. The push notification in the app arrives even when Outlook is silent. More on assessing suppliers is in the article Supplier risk in SMEs.
A second channel for the day without Teams
When Outlook and Teams fail at the same time, the very channel you would normally use to coordinate is missing. This time, according to Microsoft, Teams chats and meetings kept working for most of the outage, because Teams could switch to other gateways. You should not count on that. A list of phone numbers in the emergency folder, a messenger such as Threema or Signal for the crisis group and an agreed place where management shares updates cost little and work immediately. What matters is to try these channels out once beforehand.
CyberTell's War Room does not run on Microsoft 365. Invitations and new messages arrive as push notifications in the app, and coordination between IT, management and the IT partner stays documented in one place. If your employees sign in to CyberTell with a passkey, access does not depend on Microsoft either.
Two more measures belong in every tenant: send alerts from the Service Health view to an address outside Exchange Online, for example to your IT partner, and set up two emergency admin accounts following Microsoft's guidance. These accounts do not help against a server outage like this one, but they do help in case a sign-in error locks out all admins.
Rehearse the outage before it happens
An emergency plan that nobody knows will not help on Friday morning. A one-page business impact analysis is enough to start: five to ten important processes, and for each process the Microsoft service it depends on, how long it may be down, a manual workaround and a responsible person.
-
Record dependencies
Which processes depend on email, Teams, SharePoint or sign-in? Invoices, orders, customer service and alarms from equipment are typical candidates.
-
Define workarounds
Phone list, messenger group, printed contact details and a template for informing customers. Email is not suitable as an alarm channel for machines and equipment.
-
Write a playbook
In AI Playbooks and Tabletop, a playbook records who decides what: check the situation, inform internally, inform customers, and after the outage check the logs for phishing after-effects.
-
Practise at the table
The tabletop scenario System outage plays through an outage with an unclear cause. Part of the exercise is the question of whether it is an error at the provider or an attack.
-
Clarify reporting obligations
A pure outage at the provider is not a cyberattack and does not have to be reported to the Federal Office for Cybersecurity (BACS). If an account is taken over by phishing during the outage, things may look different.
Outage or attack? And when you have to report
In the first few minutes it is often unclear whether a provider has failed or whether someone is inside your own tenant. A rule of thumb helps: an outage hits many organisations at the same time and is visible on the provider's status page and in the media. An attack hits your own tenant and shows up in unfamiliar sign-ins, new mailbox rules or changed admin accounts. During the outage Microsoft also advised against making hasty changes to network, identity or DNS, and itself stopped all ongoing deployments. That belongs in every playbook as a principle.
Since 1 April 2025, operators of critical infrastructure must report cyberattacks to the BACS within 24 hours. A pure provider outage does not fall under this. The Data Protection Act does count the loss of availability as a breach of data security, but a report to the Federal Data Protection and Information Commissioner (FDPIC) is only required if there is likely to be a high risk for the people affected. A day without email rarely reaches this threshold. Document your assessment anyway. The article Incident response in SMEs shows how an incident runs in a structured way, from the report to the lessons learned.
Frequently asked questions
Was the Microsoft outage of November 2026 a cyberattack?
Which services were affected?
Do we get money back for the outage?
Do we have to report the outage to the BACS or the FDPIC?
How do I recognise phishing that exploits an outage?
Should SMEs move away from Microsoft because of outages like this?
Sources
- Microsoft: Preliminary Post Incident Report for Microsoft 365 (2026-11-15), copy at 365Fanatic
- CNBC: report on the Microsoft 365 outage (2026-11-12, updated until 2026-11-14)
- TechCrunch: Microsoft’s Outlook email hit by new outage (2026-11-12)
- SecurityWeek: Microsoft 365 outage disrupts Outlook, Exchange Online and Teams
- The Register: Microsoft struggling to fix massive cloud outage (2026-11-12)
- 20 Minuten: Microsoft – 74 Länder betroffen (2026-11-13)
- inside-it.ch: Microsoft-Panne schränkt Schweizer Geschäftsalltag ein (2026-11-13)
- watson.ch: So trifft die Microsoft-Störung die Schweiz
- The Guardian: How the Microsoft outage has affected key public services in the UK (2026-11-13)
- Field Effect: phishing campaigns using the Microsoft 365 outage as a lure
- Microsoft Learn: How to check Microsoft 365 service health
- Microsoft Learn: Manage emergency access accounts in Microsoft Entra ID
- Microsoft: Service Level Agreements for Online Services
- PwC Legal: ESAs publish first list of critical ICT third-party providers under DORA
- Bundesrat: Meldepflicht für Cyberangriffe auf kritische Infrastrukturen ab 1. April 2025
Third-Party Management. Assess supplier risks and ensure compliance
Everything this article describes can be put into practice in CyberTell — in the demo with sample data or in your own account.