Email Fails, Logins Break: What Really Went Down Across the Internet on November 12

How it felt on Thursday

It started with email that would not load. Then the calendar would not connect. Then a different app asked for a code by email, and the code never came.

By early afternoon on Thursday, that was the day for a lot of people. Nearly 130,000 problem reports about Outlook piled up on outage trackers just after 1 p.m. Eastern. That number counts complaints, not people, but it gives you the size of it. Teams, Microsoft's business messaging app, and the wider Microsoft 365 office tools drew thousands of complaints too.

And the trouble did not stay inside Microsoft's own products. That is the part worth slowing down for.

What Microsoft says broke

Microsoft says some of its servers dropped out unexpectedly, which pushed their work onto the servers that were left. It says the extra strain did not come from a rush of new traffic. The company paused its software rollouts, restarted affected servers and moved connections across its wider network.

The focus landed on the automatic health checks that decide whether a server is fit to work. By around 4 p.m. Eastern, Microsoft said a different way of checking had brought some servers back, and it was spreading that fix. That is recovery in progress, not an all clear for every customer. An earlier Teams problem that morning had already been fixed, so it is a mistake to treat the whole day as one long event.

Microsoft also says there is no sign a cyberattack caused this.

It reached companies that do not even run Microsoft

Here is where it gets interesting. Several other companies confirmed problems that traced back to the Microsoft failure.

Zoom said its Microsoft connections broke, including scheduling, calendar features and shared files. GitHub, the site programmers use to store their code, said some people could not sign in, sign up, verify an email or reset a password, because the emails they needed were stuck at an outside email provider. Twilio, which sends emails and texts for thousands of other businesses, said messages headed to Microsoft-run inboxes were delayed and held in line, though none were lost at the time.

None of that means those whole platforms went dark. GitHub's own site stayed up. But if the reset email never arrives, it does not matter that the website is working. You are still locked out.

The rest of the list: reported, not confirmed

Through the afternoon, complaints also climbed for Amazon, Facebook, Salesforce, Slack, Reddit, Gmail and other Google services, Amazon's cloud, the Steam gaming store and Apple's iMessage.

Some of that was real trouble. Salesforce, for example, posted a problem with one of its marketing tools on certain systems. But Slack and Reddit showed normal service when checked, Google's status page showed no matching incident late in the afternoon, and Apple's page stayed green. A green status page does not mean your connection worked. It does mean nobody has shown the whole platform failed, and we are not going to claim it did.

Other problems hit on different clocks. A power outage knocked a Los Angeles network out Wednesday evening for a content company called bunny.net, which rerouted traffic and called it fixed Thursday morning. The hosting company Fly.io had its own Los Angeles outage, and separately reset its GitHub connections Thursday afternoon. Germany's Hetzner had a few hours of storage trouble overnight. These were real, but they were not one big simultaneous collapse.

Yes, there was a real attack this week too

On Wednesday, security researchers watched rogue software spread to at least 288 computers at 48 organizations in about 90 minutes, through remote support tools that technicians use to fix other people's machines. We covered that one earlier today in our piece on hackers reaching the systems that keep us alive.

That attack is serious. But nothing we have seen ties it to Thursday's Microsoft failure or the wider wave of complaints. A bad week can have more than one cause, and the honest thing to say is that the outage is better documented than any single reason behind it.

Our take: the weak spot is the inbox

To a small business owner, none of this matters as much as one simple fact. When the login fails, the work stops. The customer in front of you does not care whose server broke.

What Thursday showed is how much quietly runs through one thing: email. It does not just carry messages. It is how you get into accounts, prove who you are and get back in when you are locked out. A second app is not much of a backup if it needs the same inbox to let you in.

So here is what we would do this week. Keep a printed or offline copy of your important contacts and this week's appointments. Have a way to reach your people that does not depend on the same email account, whether that is a phone number, a group text or a second email with a different company. If you run a business, have a plain paper way to take orders and write down sales for a few hours. And for your most important accounts, set up a backup way to sign in that is not the same inbox.

The companies involved owe their customers clear answers. What failed first, what depended on it, why recovery took as long as it did, and what changes so it does not happen again.

We do not need to invent a mystery attacker to see the lesson. When the services we lean on fail together, a lot more than one inbox stops working. The fix is the same as always: a plan that still works when the screen does not.