Microsoft's global Microsoft 365 outage on Friday, Nov 13, disrupted Outlook, Teams and Copilot for more than half a day, affecting thousands of users. The disruption spilled into Asia-Pacific business hours, according to the report.

Microsoft first blamed an "elevated load" across its infrastructure. It later said servers had unexpectedly become unavailable, which shifted demand onto the servers still running. The company is still investigating why the servers went offline and says it is not aware of any malicious activity.

Microsoft reported partial recovery in some regions after adjusting traffic protocols, redistributing traffic and introducing targeted traffic-blocking measures. The improvements were not sustained, the report says.

In Singapore, some government systems and the email services of several banks were affected. Emergency services and Singpass remained operational.

Prof Boh Wai Fong of Nanyang Technological University said companies have turned to a small number of major vendors for efficiency, economies of scale and integration. She said this can lead to over-reliance on a single vendor, so a major failure can have ripple effects worldwide. She called this a "paradox of digital transformation."

Matthew Hardman, Asia-Pacific chief technology officer at Hitachi Vantara and a former Microsoft executive, said organisations must plan for disruptions even when the failure is outside their control. "You can't outsource risk," he said. "Risk is something you have to own as an organisation."

Associate Professor Anupam Chattopadhyay of NTU said it is too early to comment on vulnerabilities without a detailed root-cause analysis. He said the complexity of cloud systems makes failures hard to identify and repair quickly, and that a quick fix may trigger other problems.

Prof Liu Yang, executive director of NTU's Cyber Security Research Centre, said "redundancy does not necessarily guarantee resilience, particularly when multiple services share underlying infrastructure or dependencies." He said better fault isolation, automated recovery and continuity mechanisms are needed.