BANGKOK – In cloud computing, where outages cost real money and speed is everything, Microsoft Azure gave its rivals a gift. A large-scale Azure outage hit yesterday, only hours before Microsoft’s quarterly earnings call, and it left businesses, gamers and remote workers struggling.
Microsoft blamed a configuration change that went wrong. The event exposed how much of the internet depends on a few giant providers. With Azure holding around 23% of the cloud market, this was no minor blip; it was a major shock.
The issue started around 9 a.m. PT (16:00 UTC) on 29 October during a standard Azure update that misfired. Microsoft traced the fault to Azure Front Door (AFD), its global content delivery and security service. That single change triggered DNS problems across Azure, which led to intermittent failures, timeouts and access errors for a wide range of services.
By midday ET, Downdetector reports surged. Azure complaints topped 18,000, rising to 30,000 when combined with Microsoft 365. Social feeds are filled with frustration. Xbox Live dropped mid-match, Minecraft worlds disappeared, and Copilot went quiet for many users.
One developer on X joked that the cloud had taken a coffee break, which summed up the mood across engineering teams.
Because the outage was not tied to a single region, the impact spread worldwide. Startups in California and large European firms felt it at the same time. Microsoft’s status page listed a growing set of affected services, including App Service, Azure SQL Database, Microsoft Entra ID, Purview and Sentinel. Even the Azure portal became unreliable, pushing administrators to fall back to PowerShell or the CLI to move traffic and keep systems running.
Who Took the Hit? Disruption Across Sectors
Consumers noticed it first through gaming. Xbox Live and Minecraft servers faltered, leaving millions staring at loading screens and switching to offline play. The timing annoyed fans. Hours before earnings, as Microsoft celebrated AI growth, the cloud tripped over itself.
Enterprises felt the bigger financial pain. Alaska Airlines, fresh from a $1.9 billion Hawaiian Airlines deal, reported outages across key systems. Its website and check-in tools stalled, echoing a separate tech failure that grounded hundreds of flights the week before.
“We’re experiencing a disruption to key systems due to the Azure outage,” the airline posted, as passengers faced booking issues. Heathrow Airport’s site went offline, Vodafone UK users saw network trouble, and delays mounted in the real world.
Retail and banking suffered too. The Starbucks app failed for mobile orders, Kroger customers hit checkout problems, and banks such as NatWest reported login issues. In the UK, Asda and M&S websites flickered, and O2 had mobile service hiccups.
Even public sector operations were touched. The Scottish Parliament linked the system’s trouble back to Azure. Cornell professor Gregory Falco told the BBC this showed the fragility of critical online services. Azure supports a large slice of the internet, and while Microsoft has not shared figures, industry watchers suggested losses in the tens of millions.
Microsoft moved quickly and shared regular updates by its standards. By 1:30 p.m. PT, engineers had isolated the faulty configuration and started a rollback to the last known good state. They expected the change to spread in about 30 minutes. Built-in safety checks inside AFD slowed this effort, which extended recovery times.
An early status update read, “We’ve failed the portal away from Azure Front Door and are assessing failover options.” Customers were advised to use Azure Traffic Manager to route around AFD while mitigation continued. By 4 p.m. ET, the pressure eased. Downdetector reports dropped below 3,300, and Microsoft aimed for full mitigation by 7:20 p.m. ET (23:20 UTC).
As of this morning, 30 October, Azure appears stable. The portal is back, though Marketplace endpoints lagged overnight. Microsoft 365 reports improved as Outlook queues cleared and Teams returned to normal. “Strong signs of improvement across regions,” Microsoft said late yesterday, and it promised a live post-incident review.
The earnings call went ahead. CEO Satya Nadella highlighted AI momentum, although some analysts on X noted the outage took some shine off Copilot’s story.
A Familiar Pattern: Azure’s Reliability Record
Azure has a history of major incidents. In 2014, a multi-region failure took down Virtual Machines and SQL Databases for hours. In 2017, power issues hit Office 365 and Xbox for more than 16 hours. In 2023, Microsoft 365 suffered a three-hour outage linked to network failures.
Recent months have been busy as well. In September 2025, Synapse Analytics jobs failed for eight hours, and Dev Box connections dropped across regions. Early October brought AFD and CDN latency issues in Africa and Europe, according to Azure’s public status log.
StatusGator counts more than 1,900 Azure incidents in the last decade. Human error in configurations or deployments ranks high, followed by hardware faults and network problems.
Some critics say Microsoft’s rapid AI expansion is stretching its capacity and operations. Last week’s AWS DNS outage, which disrupted Snapchat and Reddit, added to concerns about concentration risk. As Falco put it, many companies rely on the same three providers. When one stumbles, the wider ecosystem feels it.
The incident highlights a hard truth. The cloud brings scale and convenience, but it can centralize failure. More firms are talking about multi-cloud setups, pairing Azure with AWS or Google Cloud to build redundancy. Tools like Azure Traffic Manager will see renewed focus. Many experts also favour hybrid models that keep core systems on-premises as a safety net.
For Microsoft, the episode is a reputational knock during a strong AI run. Copilot adoption grew more than 30% last quarter, but trust drives cloud decisions. A thorough post-incident review, stricter change controls and stronger guardrails should help restore confidence.
In the meantime, teams are revisiting failover plans and traffic policies. In tech, reliability is not an optional feature; it is the baseline.
