Microsoft outage takes down Outlook, Teams and Copilot for more than 12 hours

Twelve hours of instability, across five time zones

On Thursday, November 12, 2026, Outlook, Teams and Copilot began showing connection errors across several regions of the world. More than 12 hours later, on the morning of Friday the 13th, Microsoft was still working to normalize the service: around 7:20 AM Brasília time, Teams was already responding, but Outlook and Copilot remained offline, according to a report by Exame, a Brazilian business magazine.

Downdetector, a platform that aggregates outage reports, logged around 130,000 notifications about Outlook by around 3 PM on Thursday (Brasília time). The volume dropped to 37,000 by around 8 PM, a sign of partial recovery, not full normalization.

This is the second significant Microsoft 365 outage this year, which is a data point in itself: it is not an isolated event for those who decide business continuity budgets.

The official cause: overload, not an attack

According to Microsoft, the problem lies in a portion of its infrastructure experiencing elevated load across several regions simultaneously, which generated cascading connectivity errors.

We have determined that a portion of the infrastructure is experiencing elevated load across several regions, which is likely contributing to the connectivity errors.

Microsoft, in an official statement

To contain the problem, the company paused new deployments, redirected traffic away from the affected servers and restarted compromised equipment, in a classic incident mitigation sequence: stop making it worse, divert load, restore capacity gradually.

The technical detail that matters most to those who architect systems: the failure was concentrated in resources that depend on Exchange Online, Microsoft 365's messaging backend. Tools that don't route through it remained online. It wasn't Microsoft's entire cloud that went down: it was a specific component, but a component that email, calendar, contacts and part of the notification logic for Teams and Copilot depend on.

Who felt the outage in their wallet

The list of affected parties published by Exame shows the reach of the problem beyond the obvious axis of home users:

• In Australia, the intelligence agency and the federal police operated with compromised email systems.
• In the United States, the House of Representatives confirmed that its technical teams were working alongside Microsoft.
• In Japan, the government reported impact on its own information systems infrastructure.
• OpenAI and Anthropic reported failures in the connectors that integrate their products with Microsoft services.
• In India, airlines IndiGo and Akasa Air had their email support paralyzed.
• In Taiwan, brokerages and insurers had fund management operations frozen; one company canceled an already scheduled bond buyback.
• In the United States, electric truck maker Workhorse Group postponed the release of its third-quarter earnings, from November 12 to November 16, attributing the delay directly to the Microsoft outage.

This last case turns the incident from an operational nuisance into an investor relations event: a publicly traded company moved its earnings release date because its cloud provider went down. It's the kind of line that ends up in a board meeting minute.

The parallel Australia drew with CrowdStrike

Microsoft denies any indication of an attack. Australia's Minister for Home Affairs and Cyber Security, Tony Burke, went further and compared the episode to an incident that the infrastructure sector still treats as a case study.

They believe they are dealing, in some way, with something similar to what we saw in the CrowdStrike case, where an update went wrong. The impact they are dealing with is very concentrated in email systems.

Tony Burke, Australia's Minister for Home Affairs and Cyber Security

The reference isn't gratuitous. In July 2024, a malformed update to a CrowdStrike security sensor crashed millions of Windows machines at airports, hospitals and banks around the world, a failure that had nothing malicious about it and everything operational. The pattern repeats here: it's not an intrusion, it's an infrastructure change that went wrong on a planetary scale, because the provider serves everyone through the same backbone.

The questions worth raising with the CTO this week

A 12-hour outage on Exchange Online isn't reason to switch providers overnight, but it's reason enough for the board to ask, with numbers in hand, what happens next time. Five questions make sense on that agenda:

1. Which of the company's mission-critical systems depend exclusively on Exchange Online, and is there an alternative communication channel (SMS, WhatsApp Business, a channel outside Microsoft) for the first minutes of an incident?
2. Does the Enterprise Agreement contract with Microsoft include an availability SLA for Microsoft 365 and, if it does, what is the actual financial credit for an outage of this scale, compared to the cost of lost revenue?
3. Does the monitoring team find out about a failure through Microsoft's official status page or through its own observability (synthetic checks, Downdetector, user alerts)? The difference of minutes between the two sources already decides the response time.
4. What internal process exists for decisions like Workhorse Group's, postponing an earnings release because the cloud provider went down, without it turning into an investor communication crisis?
5. What is the board's risk appetite for paying for multi-vendor redundancy (a second email and collaboration provider, even if underused) versus accepting concentration risk and paying the cost when it materializes?

The trade-off nobody wants to foot

None of these questions has an obvious answer, and that's where the work of those who decide architecture and budget together lies. Multiplying email and collaboration providers costs doubled licensing, team training on two tools and integration complexity that most companies dismiss until the day the outage makes the news.

The alternative, keeping everything with a single provider because it's cheaper and simpler to operate, is a rational bet most of the time. The problem is that this bet only becomes visible when it loses, and when it loses, it loses for everyone at the same time: there is no zero correlation between customers of the same infrastructure backbone.

In summary: the Microsoft outage doesn't change anyone's architecture overnight, but it does change the weight of a question that should already be on the business continuity agenda: how much is an hour without corporate email worth, in revenue and reputation, and who on the risk committee owns that answer?