Why Does Microsoft Outlook Keep Going Down?

How did work on a router in Madrid knock Outlook offline?

On January 25, 2023, a Microsoft engineer was adding network capacity in Madrid. A command that would have affected only one router on some equipment behaved differently on the model being used. It told routers across Microsoft’s internal network to recalculate how to reach one another. The engineer ran the same operation on a second router 33 minutes later, triggering another wave of disruption.

Outlook and Teams users reported problems as Microsoft’s network struggled to move traffic. That meant a working internet connection at your desk could still leave you unable to reach Microsoft’s services. Most regions and services recovered within about two hours, but some traffic took considerably longer to return to normal.

What broke in the October 2025 Microsoft outage?

This time the failure was in Azure Front Door, a service that receives internet requests and directs them to the Microsoft or customer systems that can answer them. An incompatible set of configuration instructions passed through Microsoft’s safeguards. The failure surfaced later, when the system processed those instructions, and users began seeing connection errors. Microsoft 365 was among the services affected.

That put the 2025 failure at a different point in the journey. The Madrid incident disrupted the routes traffic took across Microsoft’s network. The Azure Front Door incident disrupted a service that decides where incoming requests should go. In either case, the user could open Outlook and still fail to reach the service behind it.

Was the November 12 Outlook outage a repeat?

Microsoft’s early account pointed to a shortage of working servers. A number became unavailable after failing automated health checks, leaving fewer machines to handle requests. Microsoft said the pressure came from lost capacity. It also examined whether reversing recent configuration changes would help. The company had yet to establish what set off the failure in the initial updates.

Why do all three look like an Outlook problem?

Outlook is the app people see. For Microsoft 365 work accounts, reaching the inbox also depends on Exchange Online, the service that runs the mailbox, and on the systems that carry requests to it. A breakdown along that route can produce the same spinning screen as a failure in the servers at the other end. In November, Microsoft reported problems reaching Exchange Online mailboxes through multiple connection methods.

How can one change affect so many customers?

The Madrid router was one device, but it could send instructions to other routers on Microsoft’s network. In 2025, Azure Front Door was built to distribute configuration changes across many locations. In each case, a system designed to share information quickly also allowed the effects of a bad change to travel. Microsoft’s reviews show how the safeguards failed in different ways. The revised Madrid procedure had skipped retesting. In 2025, checks let the incompatible instructions through before the failure appeared.

Why can’t Microsoft just switch to spare servers?

It can move requests to other machines or use alternate routes, and it did both during the November outage. The question is how much traffic the remaining systems can handle while Microsoft brings capacity back. Some restarted servers stopped responding again when requests returned. A machine can pass an initial check and still fail when customers start reconnecting.

Why did some users in India wait longer in 2023?

Microsoft’s review singled out traffic passing through India and parts of North America as among the slowest to recover. Some routes continued dropping data after the wider network had largely recovered. Microsoft had paused automated systems that optimise traffic routes during the incident, and those systems had to be restarted manually. Indian users reported trouble with Outlook and Teams that afternoon.

What can Microsoft change to prevent another outage?

The 2023 review proposed auditing and blocking commands that could affect routers across the network. After the 2025 failures, Microsoft focused on how it tests configuration changes before they spread. Both reviews point to the same practical question for a company that runs email for businesses around the world. If a change fails on one router or at one cloud location, how much of the service can it take down with it?