Microsoft said Friday morning that Outlook and other Microsoft 365 services were beginning to recover from a widespread outage, though users could still have trouble signing in, opening their inboxes or receiving messages.

The disruption began shortly before 1 p.m. Eastern on Thursday and continued into Friday, affecting Outlook on the web and desktop, as well as Teams, SharePoint and Copilot. Some outside services that connect to Microsoft products, including Outlook Calendar, also experienced problems.

In an update shortly before 8 a.m. Eastern on Friday, Microsoft reported broader improvement in North America and sustained gains across much of Europe. The company said it was seeing encouraging signs in the Asia-Pacific region, where the outage had already disrupted much of the workday. Microsoft warned that service could remain spotty as engineers brought more systems back online.

Getting people back into their inboxes was only part of the job. Microsoft also had a backlog of undelivered email to clear. The company extended the period during which its systems keep trying to deliver delayed messages to 48 hours, giving queued emails more time to reach their recipients.

Microsoft said it had found no evidence of malicious activity. Its initial account pointed to a number of servers becoming unexpectedly unavailable, putting extra pressure on those still running. Some servers were also failing automated health checks. By late Thursday afternoon, engineers had tested a different way to check server health and used it to restore some capacity. They were rolling out that change more widely and preparing a system to detect and restart unresponsive servers automatically.

The effects reached beyond inboxes and meeting invitations. Parts of Britain’s National Health Service reported disruptions, though patients were told to keep their appointments. One hospital group warned callers to expect longer waits. Swiss luxury goods company Richemont’s earnings release was briefly delayed, while electric truck maker Workhorse Group postponed its results from Nov. 12 to Nov. 16, citing the outage. Air India rerouted incoming customer email and said its flights were unaffected.

Amazon Web Services pushed back on reports that it was also down, saying its own services were operating normally and that an unrelated internet disruption had led to inaccurate reports.

The outage showed how quickly a problem at one provider can spill into businesses that rely on the same tools for email, calendars, meetings and customer service. For Microsoft’s customers, the immediate test was whether those services would stay up long enough for the workday to return to normal—and whether Thursday’s email would finally arrive.