- 17
- November
Microsoft 365 went down worldwide for nearly 24 hours on November 12-13, 2026. Outlook email was disrupted for many organizations through a full working day. Microsoft closed the incident on November 14 and published a preliminary report on the cause on November 15. Organizations still have two jobs. The first is to clear up what this incident left behind. The second is to check whether the disaster recovery plan covers the case where the cloud provider itself fails. This article sets out the facts, the cause, what to do during an outage, and the questions executives should put to their IT team.
In one line: The incident ran from 16:17 UTC on November 12, 2026 to 16:10 UTC on November 13. It started in the network gateway in front of Exchange Online, so it hit email, calendars and the email-dependent features of Teams, SharePoint, Copilot and Defender XDR. Microsoft says it was not a cyberattack. Queued email was not lost and was delivered after recovery. Organizations should have a backup communication channel, claim their service credit, and review their business continuity plan (BCP).
What happened: the confirmed timeline
The facts here were checked on November 17, 2026. Microsoft has closed the incident. Its final report is due on November 20, so details of the cause may still change. Administrators can read the official record in Service health in the Microsoft 365 admin center, under incident IDs MO1550821 and EX1550543.
The trouble started in Exchange Online, the email service of Microsoft 365. Users in the Americas lost access to their mailboxes first. Within about an hour Downdetector had reports from 74 countries and territories. At the peak it received 19,650 reports per minute, against fewer than 15 per minute on a normal Thursday (figures from NBC News).
| Time (UTC) | Event |
|---|---|
| Nov 12, 16:17 | Exchange Online starts failing (EX1550543). Users cannot open mailboxes. |
| Nov 12, 18:26 | Microsoft opens MO1550821 for services that depend on Exchange Online and says users in any region may be affected. |
| Nov 12, about 19:06 | Microsoft pauses all deployments and configuration changes. |
| Nov 12, about 21:08 | Some servers return to service. |
| Nov 13, 00:00 | The restored servers stop responding again under real traffic. |
| Nov 13, 07:30 | Microsoft applies a patch to part of its infrastructure and reports better server stability. |
| Nov 13, 11:00 | Recovery actions complete in parts of Asia Pacific. Europe and North America improve gradually. |
| Nov 13, 16:10 | The time Microsoft later recorded as the end of the main impact. Queued email kept draining for several more hours. |
| Nov 13, 17:18 | Microsoft says the incident came from an issue during an earlier network capacity upgrade, and was not related to a cyberattack or to AI. |
| Nov 13, 20:05 | Microsoft says it is draining queued messages and rebalancing traffic, and that it does "not have an ETA for full recovery yet." |
| Nov 14, 04:17 | Microsoft closes the incident and confirms it was a technical issue, not related to a cyberattack or to AI. |
| Nov 15 | Microsoft publishes its preliminary post-incident report naming three factors. The final report is due on November 20. |
Microsoft's preliminary Post-Incident Report (PIR), dated November 15, 2026 and summarized by Wccftech, places the failure in the networking gateway servers that accept connections in front of Exchange Online. It names three factors that came together.
- A hidden defect from a Windows update. The update reached the gateway servers in July 2026. Under certain workloads an internal process ran into lock contention and used up CPU and memory. Microsoft's test labs used hardware with fewer CPU cores than production, so the defect went unnoticed for more than 3 months.
- A Teams traffic move. About 15 hours before the outage, Microsoft moved Teams traffic onto the primary gateway it shares with Exchange Online, which raised the load on that gateway.
- Capacity already reduced. About 14% of the gateway's server capacity was out of service because of an unrelated issue the day before.
Within 2 hours, more than 80% of the gateway servers had stopped responding. Traffic management sent requests to the remaining servers, which were overloaded in turn. A rollback did not help, because the defective change had been in place for months. Microsoft says the defect was specific to its Microsoft 365 gateway infrastructure and does not affect Windows deployments in general. Each of the three factors is ordinary in systems operations. The large outage came from all three landing on the same day.
Good news for users: Microsoft extended its email timeout window to 48 hours. Mail sent during the outage sat in a queue and was delivered after service returned; it was not lost. Features that do not depend on Exchange Online kept working, according to Microsoft.
Who was affected, and how they coped
Microsoft has not said how many users were affected. The Associated Press cites 450 million paid Microsoft 365 customers as the total user base, not the number that went down. Examples from international and Thai news show that the damage varied a great deal from one organization to the next.
| Group | What happened | Workaround / what kept working |
|---|---|---|
| Finance | A Taiwanese brokerage could not confirm trades by email. Euronext's listings team had trouble communicating by email. | Went back to fax and telephone |
| Listed companies | Richemont delayed its half-year results. Workhorse Group moved its earnings report to November 16. Aviva was affected around its results briefing. | Rescheduled. Aviva used a contingency plan for its results briefing. |
| Aviation and logistics | IndiGo, Akasa Air and Air India saw email-based customer service disrupted. Maersk warned customers of delays. | Flights ran normally, because operational systems do not rely on email. |
| Government and health | Government bodies in Japan, Norway, Singapore and the United Kingdom lost email. The UK National Health Service (NHS) saw delays in prescriptions reaching pharmacies. Some UK police forces were affected. | Some UK police forces moved crime reporting to the phone. |
| AI providers | OpenAI and Anthropic reported problems with their connectors to Microsoft systems. | The AI services stayed up. Only the parts that read Microsoft 365 data were affected. |
| Thailand | The Stock Exchange of Thailand (SET) said only incoming email was affected. | Trading and settlement systems ran normally. |
The table gives one clear lesson. Where core systems were separate from email, core work continued and only a communication channel was lost. Where email was a step in the work itself (confirming a trade, sending a prescription, approving a document), that work stopped with the email. The chair of the Federation of Thai Industries (FTI) made a similar point: businesses should have a BCP, a backup communication channel, and regular backups of important data.
What to do during an outage, and what is still open afterwards
The table shows what an organization should do while the service is down, split by role. Keep it for the next incident. For this week, 3 jobs are still open: check for duplicate emails and documents, confirm with customers and partners that pending items arrived, and collect the evidence for a service credit claim.
| Owner | Action | Why |
|---|---|---|
| Executives | Announce one backup channel for the whole organization, such as phone or a designated chat group, and name the work that must continue today. | Without an announcement each team picks its own channel, and work information scatters beyond recovery. |
| IT team | Follow MO1550821 in the admin center. Record start and end times, the number of users affected, and the steps taken. | This is the evidence for a service credit claim and for the review afterwards. |
| IT team | Do not mass-reset passwords or change settings to "try a fix". | The fault is on the provider's side. Changes on your side do not help and cause new problems when service returns. |
| All staff | Do not resend the same email. For urgent matters, call and say the email will follow. | Queued mail will arrive by itself. Resending creates duplicates, such as two purchase orders. |
| Sales and customer service | Tell customers and partners who are waiting that email is disrupted, and give a phone number. | Customers do not know your email failed. They only see silence. |
| Accounting and procurement | List items due today and tomorrow, such as quotations, payment confirmations and bid documents, and confirm them by phone. | Contract deadlines do not pause for an email outage. |
Watch for scam email: Security researchers reported phishing messages that used the outage as a pretext for a "password reset" within hours of the incident. Tell staff the organization will not send password reset links because of this outage, and that they should not click links in emails or chats claiming to be from Microsoft. For the wider picture, see cyber risk as the top business risk in Thailand, 2026.
How far the service agreement protects you
The Microsoft 365 Service Level Agreement (SLA) commits to 99.9% monthly uptime for Exchange Online, Teams and SharePoint. Below that, customers can claim a service credit. The table below is summarized from LazyAdmin's write-up. Check it against your own agreement before filing. This outage lasted nearly 24 hours, so users who were down throughout fall into the second row by LazyAdmin's estimate. The actual figure is calculated from the minutes your own users were unable to work.
| Monthly uptime | Approximate downtime | Service credit |
|---|---|---|
| Below 99.9% | More than 43 minutes | 25% |
| Below 99% | More than 7.2 hours | 50% |
| Below 95% | More than 36 hours | 100% |
Executives should know 4 things about this credit.
- You have to claim it. Microsoft does not apply it automatically. Open a service request in the admin center with the incident IDs, the time window and the number of users affected. The deadline is the end of the month after the incident: December 31, 2026.
- If you bought through a partner, claim through the partner. Customers of a Cloud Solution Provider (CSP) should ask their seller first.
- It is a discount on a future invoice, not a refund, and it covers only the share of the price for the service that failed.
- It does not cover business losses. A missed bid, lost sales or penalties under your own customer contracts are all outside the SLA.
The last point is why an SLA cannot stand in for a backup plan. The provider is liable up to the fee you pay. The business loss is entirely yours.
5 questions executives should ask IT after the incident
- Which of our work stopped because email was down? Have every department write the real list from this incident. It maps your weak points more accurately than a paper assessment.
- How long can each piece of work stay stopped? This is the Recovery Time Objective (RTO). How much recent data you can afford to lose is the Recovery Point Objective (RPO). Payroll and a newsletter do not have the same values.
- Does our backup channel depend on the same provider? If email, chat, file storage and sign-in all sit with one vendor, one incident can close every channel.
- If we cannot reach Microsoft 365, where is our contact list for staff, key customers and key suppliers? Many organizations found the address book was in Outlook too.
- Have we ever rehearsed? A plan that has never been rehearsed usually fails in a real incident. See the restore drill checklist for how to run one.
The long-term plan: remove single points of failure
Reuters notes this is the second major Microsoft 365 outage this year, after one in January, and that Azure had an outage in October. An eToro analyst said repeated instability could push companies to reassess relying on a single provider, while noting that large enterprises cannot move off the platform overnight.
Leaving the cloud is therefore not the answer for most organizations. Systems hosted in your own data center fail as well. What works is to decide which things need a fallback and to invest in order of importance.
| Measure | What to do | Trade-off |
|---|---|---|
| Backup communication channel | Name a second channel that is not with the same provider as email, plus an emergency contact list kept separately. | Low cost. The list needs regular updates. |
| Take approvals out of email | Approve purchase requests, purchase orders and payments inside the core business system. Email only notifies. | Needs process change and user training. |
| Separate backup of Microsoft 365 data | Back up mailboxes and key files outside the main provider. | A monthly cost. Restores must be tested. |
| Written manual procedures | For work that cannot stop, keep forms and manual steps, and a way to enter records afterwards. | Takes time to write, but no extra spending. |
| A drill once a year | Simulate "no email for one day" and measure which work gets stuck. | Disrupts routine work for half a day. |
The second measure does the most for back-office work. When purchase requests, approvals and postings already live in an ERP (Enterprise Resource Planning) system, accounting and procurement keep working on the day email fails. The same rule applies to the ERP itself: on-premise or cloud, it needs a fallback plan and a tested restore. See On-Premise vs Cloud for the comparison, and Vendor Lock-in for a view on depending on one vendor.
A caution: Do not decide to migrate the whole organization in the week of the incident. Wait for Microsoft's final report, due on November 20, 2026, then put the risk through your normal risk register process. See risk management for the assessment method.
Conclusion
This Microsoft 365 outage lasted nearly 24 hours. It was not an attack, and email was not lost. The preliminary report points to a defect that sat hidden for months and then met a traffic move and reduced capacity on the same day. This week, check for duplicate documents, file your service credit claim by December 31, 2026, and use the list of work that really stopped to update your BCP. It is data from a real incident, not from a simulation.
A cloud provider is liable up to the fee you pay. The stopped work, the waiting customers and the missed deadlines need a plan of your own.
- Saeree ERP team perspective
References
All sources checked on November 17, 2026.
- BleepingComputer — Massive Microsoft 365 outage disrupts Outlook, Teams, Copilot worldwide
- NBC News — Live updates: Microsoft outage hits Outlook, Teams across the world
- LazyAdmin — Microsoft 365 Outage: Current Status and SLA Credits
- POC Network — Microsoft's outage dragged into Friday
- Windows Central — Microsoft apologizes for major outage
- Thansettakij (Thai) — Behind the global Microsoft 365 outage: no sign of a hack
- Nation TV (Thai, citing Reuters) — Microsoft down for more than 12 hours
- Daily News (Thai) — SET confirms trading and settlement systems are normal
- InfoQuest (Thai) — FTI urges businesses to prepare a BCP
- Wccftech — Microsoft Reveals Root Cause Of Massive Microsoft 365 Outage (summary of Microsoft's preliminary report)
- Cloud Admin Hub — MO1550821 incident record
Keep the back office running, even when email is down
Saeree ERP keeps purchase requests, approvals, payments and accounting in one system, on-premise or in the cloud. Talk to our team about fallback plans and system recovery.
Request a Free DemoTel 02-347-7730 | sale@grandlinux.com



