Millions of people started Friday the same way they ended Thursday: staring at an inbox that would not load. A Microsoft failure that knocked Outlook offline Thursday morning was still lingering into a second day, with the company saying its repairs were working but slowly, one batch of servers at a time.

Early Friday Microsoft said recovery was moving forward in stages across the systems behind Microsoft 365. The company reported better results for opening mailboxes and for clearing a backlog of stuck email, yet it warned that its server capacity was still below normal and that some people would keep seeing hiccups through the morning.

Problem reports tell the story of the scale. Complaints about Outlook on the tracking site Downdetector spiked to roughly 130,000 around midday Central time on Thursday, with tens of thousands more about Microsoft 365 as a whole, from users in 74 countries and territories. By dawn Friday the Outlook count had fallen to about 15,000, a big drop but far from zero.

What broke and what still works

The trouble began at about 10:15 a.m. Central on Thursday inside Exchange Online, the cloud service that runs Outlook email for businesses, schools and governments. People could not reach their mail on the web, on desktop computers or on phones. Within a couple of hours Microsoft said users anywhere in the world could be affected.

The damage spread to tools that lean on that email system. Calendar features in Teams, such as booking or changing a meeting, often failed even though Teams calls, chats and video meetings mostly kept running. The Copilot assistant could not pull up email or files for many users. A security feature that checks links clicked inside Outlook and Teams was also hobbled.

Microsoft says a large share of its mail servers suddenly went unavailable, which piled extra traffic onto the machines that were left. A process that sets up secure connections then became swamped and those servers stopped answering too. Engineers rolled back a recent network change that lined up with the start of the trouble, restarted hardware and added virtual machines for more capacity. Late Thursday the company admitted that an earlier fix had not worked as hoped. Officials who spoke with Microsoft said it saw no sign of a cyberattack.

Cities, campuses and hospitals go old school

Across the country public offices fell back on phones, texts and social media. Michigan's state government email went down along with messages from its transportation department. The City of Detroit asked residents to expect delays on anything sent by email. Wayne State University and Monroe Public Schools reported problems too.

In New Mexico state agencies, the City of Albuquerque and the University of New Mexico all lost Outlook for much of Thursday. Classes went on as usual. A Presbyterian hospital kept caring for patients using backup channels. The state's main electric and gas utilities did the same thing. Staff in the Albuquerque mayor's office joked that they had gone analog. In Southern California the Orange County Department of Education told employees to pick up the phone for anything urgent.

The outage reached well beyond the United States. Government offices in Japan and Australia reported disruption. In Britain health workers turned to paper records while some police forces could not take crime reports online. Two of the biggest artificial intelligence companies were also caught up in it.

For anyone still waiting, information technology teams have the same advice. Restarting a computer or signing in again over and over will not help because the problem sits on Microsoft's side. Mail that was stuck is being held in a queue and should show up late rather than vanish once service is fully restored.