On a Friday that began in Australia, when people opened their inboxes to find them empty, a fault in Microsoft's cloud infrastructure reminded the wider world how much of ordinary working life now rests on systems few people ever see. What began as a strain on overloaded servers spread across dozens of countries, reaching email, shared calendars, the Copilot assistant and, through them, public counters and hospital paperwork. Microsoft insists that no attacker is behind it, and its own repairs have so far proved slower than hoped, so the day's disruption stands as a quiet lesson in dependence:
Microsoft Outage Fixes Not Going 'As Expected' as Email, Chat and AI Tools Stay Disrupted Worldwide
Workers lost a business day and some Australian public services and hospital administration were disrupted, though hospitals say patient care was not affected.
Companies have practically "lost a business day"
Mark
Why does an email outage matter this much?
Mimi
Because email is how other things get verified. Troy Hunt pointed to appointments, purchases needing verification, and account access, not just office work.
Mark
And the public services?
Mimi
Service NSW asked people to delay visits and use online options. The Northern Territory reported services down, per AAP.
Luke
Careful with scale. Downdetector's near 140,000 peak counts user reports, not confirmed affected users.
Mark
What is Microsoft saying caused it?
Mimi
Elevated load, from servers becoming unexpectedly unavailable, not from more traffic.
Luke
That's Microsoft's own account, and it's still checking what triggers the relapses. The root cause isn't established.
Mark
Is it a cyber-attack?
Mimi
Microsoft doesn't think so, according to Tony Burke, who compared it to CrowdStrike.
Luke
That's a minister relaying what Microsoft officials told him. And Paul Spain's "incompetence" is an expert's opinion, not a finding.
Mark
Any sign of patient harm?
Mimi
Hospitals report clerical issues only, and the AMA's Michael Bonning says workarounds exist.
Luke
That's their statement. Nobody has given a recovery time yet.
The Pulse
- Outlook, Exchange Online and Copilot went dark for tens of thousands of users, with reports on Downdetector peaking near 140,000 worldwide.
- Within about an hour the trouble had reached 74 countries, beginning in Australia and then rolling across time zones.
- Service NSW asked customers to postpone visits and finish their business online, while Northern Territory services stalled.
- Microsoft blames overloaded servers and says its first fixes are not taking hold, so restarted systems may fail again under traffic.
- Officials say this is not a cyber attack but a faulty change akin to the CrowdStrike failure, and businesses are counting the cost of a lost working day.
On a Friday that began in Australia, when people opened their inboxes to find them empty, a fault in Microsoft's cloud infrastructure reminded the wider world how much of ordinary working life now rests on systems few people ever see. What began as a strain on overloaded servers spread across dozens of countries, reaching email, shared calendars, the Copilot assistant and, through them, public counters and hospital paperwork. Microsoft insists that no attacker is behind it, and its own repairs have so far proved slower than hoped, so the day's disruption stands as a quiet lesson in dependence: when one shared foundation falters, many lives feel the tremor at once.
In Australia, one of the first places to wake on Friday, people opened their email and found nothing. Service NSW, which handles government transactions across New South Wales, asked customers to delay visits to its centres and finish online where possible. Northern Territory services were also down. The trouble had been building across Microsoft's cloud since the previous afternoon, and by the time it peaked it had reached 74 countries.
The core of the disruption lies in Microsoft 365. Exchange Online, which carries the Outlook app, was unavailable for many customers, some could not see certain files or calendar entries, and Copilot, the company's AI chatbot, faltered as well. Microsoft says it has blamed an "elevated load" on infrastructure in several regions. Servers became unexpectedly unavailable, and the surviving ones were overwhelmed. Restarted services may become unresponsive again once traffic returns, and the company is still investigating what set this off.
Repairs have not gone smoothly. Microsoft said its changes were "not taking as expected" and is testing other ways to let users back in without overloading its systems, though it later reported positive signs as systems restarted. Australia's Home Affairs and Cybersecurity Minister Tony Burke said Microsoft does not believe a cyber event is involved, and compared the failure to the faulty CrowdStrike update of July 2024. Gorilla Technology's Paul Spain called the situation a serious lapse that could cost millions if it lingers, and said companies had effectively lost a business day.
The effects reached beyond Microsoft itself. OpenAI and Anthropic both reported problems with their links to Microsoft services. Some Australian hospitals reported clerical disruption but said patient care was unaffected, and the Australian Medical Association noted that such systems are not used for direct care and that workarounds exist, with some facilities falling back on paper for scripts. Troy Hunt observed how central email has become to appointments, verifications and account access, and urged businesses to plan for such days. Online, people found a little humour in the chaos, and Microsoft has offered no timeline for full recovery.
In Australia, one of the first places to wake up to it on Friday, people opened their email and found nothing. Service NSW, the one-stop shop for government services in New South Wales, which includes Sydney, told customers to put off visits to its service centres and finish their transactions online where they could. Authorities in the Northern Territory reported their services were down, according to the Australian Associated Press. The cause was a disruption across Microsoft's cloud services that had been building since the previous afternoon.
Downtime-tracking site Downdetector says the first problems appeared just after 16:15 GMT on Thursday. Within about an hour the outage had reached 74 countries. Reports of Outlook trouble peaked at nearly 140,000 around 18:00 GMT, more than 2,000 times the normal level. By the time of the latest update the site counted more than 28,000 affected users, and it said the problem had run for at least eight hours.
The disruption centres on Microsoft 365. Exchange Online, which includes the Outlook email app, was down for many customers. Some users could not view certain files or calendar details, and Copilot, Microsoft's AI chatbot, was disrupted too. Microsoft said messaging and online meetings were largely working. Its website says more than 430 million people use Microsoft 365 apps, and it says more than 90% of Fortune 500 companies use Copilot. Tracking platforms put the number of companies on Microsoft 365 at between three and four million.
Microsoft has blamed "elevated load" on its infrastructure in multiple regions. It said this was not a surge in traffic but a number of servers being "unexpectedly unavailable", which strained the rest of the system. It later said a key part of the affected servers was "becoming overwhelmed". It also warned that services that had restarted could become unresponsive again once traffic returned, and that it was checking what triggered this.
The repair has not gone smoothly. Microsoft said its changes were "not taking as expected" and that it was testing other approaches to let more users back on without overloading its systems. A later update on X said the team was "seeing positive results" restarting affected systems and was monitoring how they performed.
Australia's Home Affairs and Cybersecurity Minister Tony Burke said he had spoken to Microsoft officials in the country, who "do not believe they are dealing with a cyber event". He said they saw something similar to the CrowdStrike failure, where an update went awry. That faulty update, in July 2024, froze Windows computers for millions of people. Burke said the impact was mostly on email systems, though CrowdStrike had brought downstream effects on many other products. Paul Spain of New Zealand advisory firm Gorilla Technology said that if it was not an attack, it reflected "incredible incompetence" that could cost millions of dollars if it dragged on. He said companies had effectively "lost a business day".
The effects spread beyond Microsoft's own products. OpenAI said it was investigating an issue with its connections to Microsoft's services, which left some tools not working as expected. Anthropic said it had found a problem affecting its links to Microsoft 365. Some Australian hospitals and healthcare bodies reported clerical problems but said patient care was unaffected. Michael Bonning, president of the Australian Medical Association, told The Guardian that none of these systems are used for direct patient care and that workarounds always exist. Some of the association's facilities had gone back to paper for scripts.
Cyber security expert Troy Hunt, based in Australia, said Microsoft is a huge part of the internet's backbone and that the outage showed how critical email remains. He pointed to appointments, purchases needing verification, and account access as things that depend on it, and said businesses need contingency plans. Meanwhile, users turned to jokes online, including a post about sending emails through Excel in the manner of Kelly Rowland in the video for Dilemma, which drew 23,000 likes. Microsoft has given no time for full recovery.
Notable Quotes
Microsoft officials do not believe they are dealing with a cyber event.— Tony Burke, Australia's Home Affairs and Cybersecurity Minister
Microsoft is a massive part of the internet backbone and reminds us how critical email still is.— Troy Hunt, cyber security expert
Want the full story?
Read the original at BBC ↗