Microsoft has announced that a portion of its servers are returning to operation and that a simplified recovery process has been initiated following a global outage of Microsoft 365 services. According to company data, nearly 130,000 users worldwide reported difficulties accessing the software suite, including Outlook and Teams.
The outage tracking website Down Detector recorded a sharp spike in reports of issues with Microsoft 365. Nearly 130,000 reports related to restricted access to the Word, Excel, PowerPoint, Outlook and Teams applications. Difficulties were also noted in business services such as SharePoint and OneDrive, and the AI assistant Copilot was affected. Many users stated that they were unable to send or receive email or experienced issues loading their inboxes. More recent reports have decreased since that peak.
Microsoft posted on the social network X that it was investigating reports from users who had difficulties accessing Exchange Online mailboxes via multiple connection methods. The company explained that part of its infrastructure was under increased load, which likely contributed to connection errors. It was also determined that a certain number of servers had unexpectedly become unavailable due to an issue that caused them to fail during automated checks.
Microsoft stated that it had confirmed an alternative health check method on a subset of the affected servers, enabling their successful return to operation. The company noted that this fix is being applied more broadly and that it is preparing an automated process to detect and restart unresponsive servers. A review of recent changes identified configuration upgrades as a possible cause of the problem, although the investigation was still ongoing.
Microsoft reported that it was working to restart and validate the affected servers and was striving to accelerate the process. The company considered whether rolling back a small number of recent configuration upgrades could speed up recovery. It continued to seek all available means to restore the service, as actions taken so far had not provided the expected relief. It was determined that a key process on the affected servers was becoming overloaded, causing the servers to stop responding.
Microsoft subsequently announced that some of the affected servers had begun returning to operation and that the others were stable. The company redirected traffic and connections to alternative paths where possible and observed limited recovery in some regions. The recovery process has been simplified to bring the remaining affected servers back into service as quickly as possible. Traffic throttling was also introduced to reduce the load on affected servers. Microsoft warned that users should expect a gradual recovery as they reconnect to 365 services.
An earlier service degradation alert on Microsoft’s pages stated that the company was aggressively seeking all ways to mitigate the impact and was restarting affected servers to bring them back into operation. As a precaution, Microsoft paused all deployments and configuration changes while reviewing recent modifications to find a correlation with the incident. It was determined that the increased load was not caused by an overall rise in traffic, but by a large number of servers unexpectedly becoming unavailable, which redirected the load to the remaining infrastructure. Microsoft investigated what was causing the server unavailability while working to safely return them to service.
Mark Gregory, an associate professor at the RMIT University School of Engineering, assessed that the outage affects the network layer in Microsoft’s systems. According to him, the outage is potentially similar to a configuration issue that took down Optus’s network in 2023. Gregory explained that when some servers have configuration problems, users are forced to share a smaller number of servers to access cloud services. The loss of some servers means that all users attempt to access the remaining ones, overloading them and causing them to fail, leading to a cascading problem. He added that it takes time to find the root problem, in this case a misconfiguration in the network layer. Companies, he believes, should invest more in test environments for configuration changes before implementation in the live network. Enterprises that rely on cloud services such as Microsoft 365 should have backup options, Gregory said. From a global perspective, someone somewhere was seriously affected and it ultimately leads to a loss of productivity, he highlighted. Many people, he added, hold a Gmail account, so if Microsoft goes down they use Gmail, and if Gmail goes down they use Microsoft.
OpenAI and Anthropic reported that the issue affected the Microsoft 365 connector that links ChatGPT and Claude AI with Microsoft’s services. Earlier, Microsoft announced that it had successfully resolved issues related to Microsoft Teams in North America.
