Microsoft has failed for more than 14 hours to fully restore Outlook, Teams and its AI assistant Copilot after a global outage, leaving thousands of workers across 74 countries facing a disrupted working day on Friday.

The company said its recovery attempts had produced only partial results, and warned that services could become unresponsive again even after being restored. Microsoft attributed the disruption to servers becoming unexpectedly unavailable, which placed additional strain on its remaining infrastructure. It ruled out a cyber attack.

"We're continuing to work on recovering our frontend servers while implementing further traffic optimisation actions to improve service stability and increase capacity," Microsoft said. "We're aggressively pursuing all pathways to reduce impact."

Reports of problems accessing Outlook peaked at nearly 140,000 globally around 6pm GMT on Thursday, according to monitoring website Downdetector. Businesses across Britain and Europe began their working day with email, meetings and internal communications still impaired.

What does the outage change for business users?

The disruption threatens to derail the working day for British businesses that depend on Microsoft's software to run email, meetings and internal channels. Banks, government agencies and courts across several countries have already been disrupted, with some organisations forced to delay operations.

One corporate casualty was immediate. Swiss luxury goods group Richemont delayed the announcement of its half-year financial results on Friday morning because of the outage, although it subsequently published the figures.

More than 430 million people worldwide use Microsoft 365 applications. The blast radius extended beyond Redmond's own customers: AI firms OpenAI and Anthropic both reported problems with services connected to Microsoft's infrastructure.

Technology analyst Paul Spain warned the disruption could become the "most severe technology outage in history" if it continued for 24 hours, potentially affecting billions of dollars in business operations.

Why one vendor can stall the global economy

The outage has sharpened the debate over single-vendor dependence. Toby Murray, a computing professor at the University of Melbourne, warned of the "inherent dangers" of such reliance, saying a major outage at one provider could "dramatically reduce business productivity globally".

For practitioners, the calculus is straightforward: when email, calendars, chat and increasingly AI tooling sit with one supplier, a single server failure can freeze client contact, internal approvals and reporting cycles at once. Richemont's delayed results announcement shows how quickly the damage reaches investor communications.

The episode also lands as European governments push to reduce their dependence on American technology providers and strengthen domestic digital infrastructure. A 14-hour blackout affecting 74 countries hands ammunition to policymakers arguing for sovereign alternatives.

Is this part of a pattern?

The incident follows a separate outage at Microsoft's Azure cloud platform last month. It also echoes the global IT failure in 2024, when a faulty CrowdStrike update grounded flights and disrupted banking services — the closest recent comparison, though that event traced to a third-party update rather than Microsoft's own infrastructure.

The timing compounds the pressure. The outage comes at a bruising moment for the US technology sector, after Nvidia shed nearly $600bn in market value on Thursday following the resignation of its long-serving chief executive Jensen Huang.

The chip giant's collapse fed a wider tech sell-off, with investors already questioning whether the billions poured into artificial intelligence will deliver sufficient returns. Microsoft sells Copilot as a flagship of that AI investment; on Friday the assistant sat dark alongside the email and chat tools it is meant to augment.

For executives, the practical takeaway is contingency: check whether fallback channels for email, meetings and investor communications exist outside Microsoft's stack, and test them before the next frontend server fails.

Microsoft has not given a timeline for full restoration. Until it does, the company's own warning stands — services restored today may drop again tomorrow.