The outage that started on 12 November 2026 particularly affected access to Exchange Online mailboxes and message delivery. Other Microsoft 365 services were also affected in some functions that depend on Exchange Online. This article looks at which continuity options worked for email, file access and local applications after the incident.
What happened: timeline of the outage
Times have been converted from the UTC values in Microsoft’s preliminary report to Türkiye time (TRT, UTC+3). The report gives the start of the incident as 12 November 16:20 UTC and the end as 13 November 16:10 UTC.
| Time (TRT) | Event |
|---|---|
| 12 November, 04:07 | Automatic failover moved Teams traffic to gateway servers shared with Exchange Online. During the day, the load on these servers increased gradually; individual server failures occurred in some regions. |
| 12 November, 19:20 | A high-priority alert was received for Exchange Online and incident EX1550543 was opened. According to the report, this is the start of the incident. |
| 12 November, 21:27 | The impact on Teams, Copilot and Defender was announced under incident MO1550821. |
| 12 November, 22:00 | All deployments and configuration changes were halted. |
| Night of 12–13 November | Targeted changes, restarts and alternative health checks were tried. Servers returned to service became unable to respond again under traffic; recovery remained limited. |
| 13 November, 10:20 | A fix reverting the defective change was applied to part of the capacity. After stability was confirmed, the fix was extended to other regions. |
| 13 November, 14:35 | The Europe region returned to healthy levels. |
| 13 November, 19:10 | Customer impact ended. The team moved on to monitoring and processing queued messages. |
The incident numbers in the Microsoft 365 admin center are MO1550821 (other services depending on Exchange Online) and EX1550543 (Exchange Online mailbox connectivity). You can see the impact on your own Microsoft 365 environment (tenant) and the full report from these two entries.
Scale of the impact
According to the preliminary report, the main impact was on Exchange Online: most users couldn’t connect to their mailbox with Outlook or other mail clients, and email delivery was delayed for many users. The parts of Copilot, Teams, SharePoint, OneDrive, Defender and Purview that use Exchange Online were affected; the report states that these services remained largely available.
What Microsoft said about the cause
The preliminary report explains the outage as the combination of three factors in the gateway layer that sits in front of many Microsoft 365 services:
- Software defect: A defect in a Windows fix prepared for Microsoft 365 gateway servers consumed processor and memory under certain load. The fix had been deployed in July and had run for more than three months without causing problems.
- Increased load: About 15 hours before the outage, Teams traffic was automatically moved to the gateway shared with Exchange Online. The load increased and the traffic pattern changed.
- Missing capacity: Because of a separate problem the day before, about 14 percent of gateway server capacity was out of service.
According to the report, the failure spread from server to server, and the isolation mechanisms that should have kept the problem in a single region didn’t work because of a separate issue. Recovery took longer because servers returned to service couldn’t handle the pending traffic. Since the defective change had been made months earlier, the usual rollback tools couldn’t be used.
Among the completed work, Microsoft lists reverting the connection handling code to a known good version, adding capacity, and new monitoring rules for traffic routing and connection health. The full list of repair items will be in the final report.
What the SLA gives, and what it doesn’t
Microsoft commits to 99.9 percent monthly uptime for Exchange Online and other online services. In a thirty-day month, that means about 43 minutes of allowed downtime. This outage, which lasted about 24 hours, is far above that allowance.
When the commitment is exceeded, the remedy is a service credit, i.e. a discount deducted from the next bill. This credit doesn’t cover the cost of the work that stopped during the outage. That’s why the continuity plan should be based not on the SLA percentage but on how many hours of outage the business can withstand.
If everything is in the cloud, what do you have left during an outage?
In an outage like this, how much of the work continues depends on which service runs where. The table below compares two common setups in this outage scenario.
| Service | Fully in the cloud | Critical components on premises |
|---|---|---|
| User sign-in and identity | Depends on the cloud identity service; if the service is affected, sign-in and application access are disrupted | Domain-joined computers sign in with the local domain controller (DC) |
| File shares | Access to SharePoint/OneDrive depends on the outage | The file server keeps working on the local network |
| ERP, accounting, production software | If SaaS, depends on the provider’s status | Local SQL and application servers are unaffected |
| DNS, DHCP, printing | Usually local already; indirect impact if cloud-managed | Unaffected |
| No access during the outage | Affected the same way if mail is in Exchange Online | |
| Data copy | Only on the provider’s platform | An independent backup copy on premises or at a separate target |
The last two rows of the table matter. If the mailbox is in Exchange Online, an on-premises server doesn’t save email. On the other hand, accounting, production and file services running on the local network aren’t directly affected by this outage; the business can keep issuing invoices and accessing its files.
Why on-premises servers still matter
This article doesn’t recommend leaving the cloud. Microsoft 365 is a suitable email and collaboration platform for many businesses. Our recommendation is not to concentrate dependency in a single point. The decision between cloud and on premises should be made service by service, not in one go.
1. Identity should stay local
In a hybrid identity setup, users are kept in on-premises Active Directory and synchronized with Entra ID. Even if there is a problem on the cloud side, signing in to computers and accessing the local file server and applications continue through the local DC. At least two DCs, healthy DNS and a backed-up synchronization server are the foundation of this setup. For synchronization options, see our Entra Connect Sync or Cloud Sync article.
2. The applications that keep the business running should run on the local network
ERP, accounting, MES, warehouse management, license servers. In many businesses, these systems stopping directly stops the work. When they run on a local virtualization cluster, they aren’t affected when the internet connection or a SaaS provider goes down. A two-node cluster instead of a single server keeps running through hardware failure too; the planning steps are in our Hyper-V Failover Cluster article. For an example of moving to this setup, see the JMW Jant VxRail to Hyper-V migration story.
3. The file server didn’t die; its role changed
Moving files entirely to SharePoint doesn’t suit every business. Large CAD files, production drawings and frequently changing shared folders open much faster on the local network. With Azure File Sync, frequently used data stays cached on the local server and a copy sits in the cloud. If the cloud is unreachable, work continues with the cached files. Details are in our hybrid file server article.
4. There should be an independent copy of the data
The preliminary report doesn’t mention data loss. But in a setup where the only copy sits on the provider’s platform, when the provider is unreachable, your data is unreachable too. Backing up Microsoft 365 data to a separate target lets you reach a critical email or document independently of the provider during an outage. The methods are in our Microsoft 365 backup article; for the number and location of copies, the 3-2-1-1-0 rule applies.
5. You stay in control
With on-premises infrastructure, you choose the maintenance window. You try the update on a test server first, then roll it out. When a problem arises, you connect to the console, read the log and intervene. With a cloud service, intervention is in the provider’s hands; the IT team monitors the service status and informs users.
On premises has its costs too
On-premises infrastructure doesn’t provide continuity on its own. In an unmaintained server room, a single disk, a single power supply, an untested backup or an operating system with delayed patches becomes a cause of outages.
- Hardware and redundancy: RAID, dual power supplies, UPS, preferably a two-node cluster.
- Monitoring and patching: Disk, service and certificate alerts must be monitored, and a monthly patch cycle applied.
- An on-site copy isn’t enough: An offsite, immutable copy is a must against fire, theft and ransomware.
- On-premises Exchange isn’t the cure for this outage: Versions of Exchange Server other than the Subscription Edition are out of support. An on-premises mail server brings a patching burden and enlarges the attack surface because it is exposed to the internet. Leaving mail in the cloud and preparing a backup plan for communication is more suitable for many businesses.
Make the decision based on targets. How many minutes can each service be down, and how many hours of loss can each piece of data tolerate? Your answers to these two questions determine where a service runs. For the method, see our choosing technology by RPO/RTO target article.
Checklist for the next outage
- Map the dependencies. Which service does each business process depend on, and which provider does that service depend on? See it in a single table.
- Define a communication channel for outages. Email and Teams are on the same platform. Have a phone list, an SMS group or a messaging channel with a different provider defined in advance. Keep the list printed and local, not in a file in the cloud.
- Keep critical applications local or deploy them redundantly. A system whose stopping means lost business shouldn’t depend on a single external service.
- Back up Microsoft 365 data to a separate target and test restores.
- Automate service status tracking. Turn on Service Health notifications in the admin center; when the admin center can’t be reached, the Microsoft 365 status page can be used.
- Write the outage procedure into a runbook. Who decides, what users are told, when to escalate? For a template, see our disaster recovery runbook article.
What the IT team should do during an outage
- First separate whether the problem is on your side or the provider’s: Service Health and incident numbers.
- If it’s caused by the provider, don’t tinker with the configuration. Changing DNS, connectors or licenses creates new problems after recovery.
- Tell users that delayed mail will arrive and that they shouldn’t resend; in this outage, Microsoft extended the delivery time so delayed messages wouldn’t bounce.
- Warn about phishing. During outages, fake emails and SMS messages themed “reset your password” can increase.
- After the outage, read the post-incident report and update your own dependency map accordingly.
Cloud services can have outages. Whether an outage stays limited to communication or stops the business is decided by the preparation done in advance.
To assess together which services in your infrastructure belong in the cloud and which on premises, see our server and cloud management, Microsoft 365 and cloud services and backup and disaster recovery pages. Designing and maintaining hybrid infrastructure is the daily work of the Eretek systems support team. For other continuity topics on the Microsoft 365 side, our continuity in Microsoft 365 and cloud infrastructure article is a follow-up to this one.
Frequently asked questions
Was the 12–13 November Microsoft 365 outage a cyberattack?
Microsoft’s preliminary post-incident report dated 15 November 2026 explains the outage as a combination of a software defect, increased load and missing capacity; it doesn’t mention an attack.
Were emails sent during the outage lost?
According to the preliminary report, email delivery was delayed for many users. Microsoft stated that it extended the delivery time so that delayed messages wouldn’t bounce as undeliverable, and that all queued messages were processed. The status of a specific message can be checked with message trace in the admin center.
Will Microsoft pay compensation for this outage?
When the monthly uptime commitment is exceeded, Microsoft’s contractual remedy is a service credit; it is claimed from the admin center within the period in the agreement. It doesn’t cover the cost of the outage to the business.
Does going back to on-premises servers prevent outages like this?
Not on its own. As long as mail stays in Exchange Online, email is affected by the outage. The benefit of on-premises servers is that identity, files and business applications keep working independently of the cloud. The right approach is to decide where each service runs based on outage tolerance.