Microsoft’s own service-health record dates the Microsoft 365 outage of 12 and 13 November at 23 hours 50 minutes, by our arithmetic, and its preliminary review says that at the lowest point more than 80% of its network gateway servers were unresponsive. CyberCube estimates the insured cost at most likely less than $100 million.
This explainer gathers the seven numbers that define the outage, each taken from Microsoft’s records, its preliminary post-incident review or the estimates and reports cited below, and each already covered in our reporting since 12 November. The Register reported that Microsoft’s final post-incident report is due on 20 November.
The Seven Numbers
1. 104 to 134 days. How long the faulty July gateway fix ran before it failed on 12 November, by our arithmetic from Microsoft’s statement that it “first started deployment in July of this year”.
2. Nearly 14% to more than 80%. The share of gateway capacity already out of service at the start, against the share of servers unresponsive at the lowest point, per Microsoft’s review.
3. 23 hours 50 minutes. Microsoft’s recorded impact window, 16:20 UTC on 12 November to 16:10 UTC on 13 November, by our arithmetic. Its public posts spanned 36 hours 15 minutes.
4. 15 hours 17 minutes. From Microsoft’s first internal alert at 16:03 UTC to the mitigation at 07:20 UTC the next day, by our arithmetic from the timings The Register reported.
5. 84 organisations in 12 countries. Our tally of named organisations that said, or were reported by BBC News, Reuters and WDR to have said, that the outage affected them; 26 were health bodies.
6. Less than $100 million. CyberCube’s most likely estimate of global cyber insured losses, with a conservative upper bound of $350 million. Kovrr put the figure at around $700 million, Reuters reported.
7. 96.7%. The November uptime of a user cut off for the whole recorded window, by our arithmetic, which falls in the band where Microsoft’s SLA pays a 50% service credit.

Business Pill · Concentration Explained
An evergreen Business Engineer explainer on concentration, when a few big pieces make the total. It teaches the general idea only and does not discuss Microsoft or this outage.
The key insight: As we read the seven numbers, the outage was long and cheap at once: Microsoft’s record counts 23 hours 50 minutes of impact, yet CyberCube’s most likely insured cost is under $100 million, because the remedy for most customers is a capped service credit and email sat outside most firms’ transaction path, as CyberCube put it.
What Happened, Hour by Hour?
The fault began months earlier. Charles Lamanna, Microsoft’s EVP of Copilot, Agents, and Platform, wrote on X on 16 November that a fix “prepared specifically for Microsoft 365 networking gateway servers” started deploying in July and that “for more than three months, the change was deployed worldwide without impact”.
On 12 November, Teams failed over to the “primary” gateway at 01:07 UTC, The Register reported, citing Microsoft’s preliminary report. A connectivity anomaly monitor flagged rising issues at 16:03 UTC and the first support case arrived at 16:33 UTC, it reported. Microsoft’s service-health record gives “Start time: Thursday, November 12, 2026, at 4:20 PM UTC”.
The Register reported that the mitigation, “an update that reverted the defective change”, was applied at 07:20 UTC on 13 November and that Microsoft confirmed health at 16:10 UTC that day. The record gives “End time: Friday, November 13, 2026, at 4:10 PM UTC”. Our root-cause piece follows Microsoft’s updates in between.
Why Did Microsoft 365 Go Down?
Lamanna wrote that “a latent defect caused servers to become overwhelmed and unresponsive under certain load conditions”, due to “lock contention inside an internal process”. Load testing missed it because it “tended to focus on our weaker hardware SKUs with fewer cores”, he wrote.
Two routine conditions met the bug. The Teams failover “increased the load on the networking gateway and changed its traffic patterns”, and “nearly 14% of server capacity for the networking gateway was out of service due to an unrelated issue from the previous day”, Lamanna wrote. Rerouted traffic then spread the fault: “At its lowest, more than 80% of servers were unresponsive”.
By our arithmetic, that is about 5.7 times the share of capacity out at the start. Our July fix piece walks through the review, and our latent bugs explainer sets it against AWS and Google Cloud precedents.

How Long Was Microsoft 365 Down?
It depends on who counts. Microsoft’s record gives 23 hours 50 minutes. From the first @MSFT365Status post at 17:54 UTC on 12 November to the post at 06:09 UTC on 14 November saying it had restored service to the remaining customers and regions is 36 hours 15 minutes, by our arithmetic.
CyberCube said severe disruption lasted about 18 hours and the event ran roughly 38 hours in total, “well beyond the eight to 12 hour waiting periods common in contingent system failure cover”. Exchange Online took longest: Lamanna wrote that Teams, OneDrive, SharePoint and Office apps failed out of the faulty gateway, “But not all services, like Exchange Online, can fail out of this network gateway”.
Who Was Hit?
Reuters reported that Downdetector logged more than 7.4 million reports worldwide, peaking at nearly 20,000 a minute, more than 1,500 times normal levels. CyberCube said users in more than 70 countries were cut off.
Our tally of 84 named organisations counts 60 in the UK, where the BBC’s regional reporting was densest, and 26 health bodies. It is a floor, not a census. Our pool alarm piece covers the German park that said the outage stopped its chlorine warning emails.
What Did the Outage Cost?
CyberCube estimates global cyber insured losses “will most likely be less than $100M, with a conservative upper bound estimated at $350M”, a loss ratio impact of 0% to 2% of global cyber premium. It said “email is not in the critical transaction path of most businesses”. By our arithmetic, under $100 million over about 18 hours of severe disruption is less than about $5.6 million an hour.
Estimates differ. Reuters reported that Kovrr’s Shalom Bublil put insured losses at around $700 million, from economic losses of about $16 billion on a lost-productivity basis, and called it a “back-of-an-envelope” assessment. Our insured-loss piece sets the four estimates side by side.
For customers, the remedy is a credit. A user cut off for all 1,430 recorded minutes ends November at about 96.7% uptime, by our arithmetic, in the band where Microsoft’s SLA pays 50% of the monthly fee. The SLA says credits “will not be awarded to compensate for any other forms of loss, including but not limited to lost revenue, operational costs, or any indirect losses”. Our SLA math piece has the tenant-level arithmetic.
What Is Microsoft Changing?
For the recovery, Lamanna wrote, engineers “developed an update that returned the code path to the previously known good state” and applied it to groups of servers, “(e.g. 50%+ of a region)”. The Register reported that all ongoing and planned deployments and configuration changes were paused while the team worked to restore service.
Lamanna wrote that “our normal rollback, flighting, and feature flags were unable to quickly revert the change” because the defect was months old, and that “Our normally automated recovery processes were not designed to restore service from this broad unresponsive state”. He wrote that “The final Post-Incident Review (PIR) will provide more details later this week”; The Register reported it is due on 20 November.
The Structural Read
As we read the review, the length came from the recovery, not the trigger: the defect reached the fleet in less than two hours, Lamanna wrote, while the mitigation came 15 hours 17 minutes after the first alert, by our arithmetic.
As we read the cost figures, they measure different things: CyberCube and Kovrr estimate insured losses, Kovrr’s $16 billion is lost productivity, and Microsoft’s SLA credit is capped at the monthly fee for the affected service.
Jon Laux, CyberCube VP of Products and Analytics, 16 November 2026
“This event once again reminds us how the technology world is concentrated in the hands of a few key providers.”
Three Implications
FOR IT ADMINS A November incident points to a 31 December claim deadline under the SLA’s own example, by our reading, and claims must include the number and location of affected users.
FOR INSURERS CyberCube put the loss ratio impact at 0% to 2% of global cyber premium, with most outcomes clustering toward the lower end.
FOR OPERATORS Lamanna wrote that load testing tended to focus on weaker hardware SKUs with fewer cores, while the contention is worsened on machines with more cores and newer hardware.
The Business Engineer Lens
This story maps onto the Business Engineer framework Enterprise AI: From Software to Substrate.
The framework argues: “enterprise vendors don’t sell software anymore—they sell irreversibility.”
As we read it, the outage shows what that irreversibility costs a customer on a bad day: Exchange Online could not fail out of the shared gateway, and the SLA remedy is capped at the monthly fee. The framework predates the outage and does not discuss it.
Our full coverage of the outage
- Microsoft 365 Root Cause: Latent Bug Behind 36-Hour Outage
- Microsoft’s July Fix Sat Latent 104-134 Days, Then 80% Froze
- Microsoft Outage Insured Losses Under $100M, CyberCube Says
- Microsoft Outage Hit 84 Named Organisations in 12 Countries
- Microsoft 365 SLA Math: A Full-Day Outage Means 96.7% Uptime
- SLA Uptime Explained: What 99.9% Allows and Credits Pay
- Latent Bugs Explained: Microsoft’s 23h50m vs an 11h Median
- Park Says Microsoft Outage Blocked Its Pool Alarm; 15 Hurt
- 3 Outages, 46 Hours in Two Days: Microsoft, Workday, the FAA
What Is Not Established
We did not read the full review in the Microsoft 365 admin center; its quotes and timings here come from Lamanna’s abridged post and The Register. The texts we read give no day in July, no count of gateway servers and no count of affected customers.
The loss figures are estimates by CyberCube and Kovrr, not claims paid, and our tally of 84 organisations counts only those named in the reports we read. Microsoft was not contacted for this piece.
The Bottom Line
By Microsoft’s account, a July gateway fix ran for 104 to 134 days, by our arithmetic, until a Teams failover and a capacity gap of nearly 14% met it, and more than 80% of servers became unresponsive. Its record counts 23 hours 50 minutes of impact; CyberCube puts most likely insured losses below $100 million. The Register reported that the final report is due on 20 November.
95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. This explainer draws on sources read in full for our earlier pieces on the outage: Charles Lamanna’s X post of 16 November 2026, an abridged version of Microsoft’s preliminary post-incident review; The Register’s report of the same day, which quotes the review and gives its timings; Microsoft’s final service-health entry for incident MO1550821 as republished by Cloud Admin Hub and PennO365; Microsoft’s October 2026 SLA; Reuters and BBC News reporting on who was hit; CyberCube’s release of 16 November; and Reuters on Kovrr’s estimate, as carried by Business Recorder.
We did not read the full review in the Microsoft 365 admin center. The day counts, elapsed times, the 5.7-times ratio, the per-hour figure, the 96.7% uptime and the tally of 84 organisations are our own arithmetic. Microsoft was not contacted for this piece.
Sources: Charles Lamanna (Microsoft EVP) on X: abridged root cause, 16 Nov 2026 · The Register: Microsoft blames monster email outage on software defect introduced in July fix (16 Nov 2026) · Microsoft 365 service health MO1550821, final entry (via Cloud Admin Hub) · Microsoft Service Level Agreement for Online Services, October 2026 · CyberCube: less than $100M global insured losses from M364 outage (16 Nov 2026) · Reuters via Business Recorder: Kovrr $700 million estimate (13 Nov 2026)









