Outage echoes year-old AWS cascade in Virginia

A power outage at a Google datacentre in London on Tuesday cascaded into service disruptions across the United Kingdom, with user-submitted reports to the website Downdetector peaking around 9am and partially resolving by mid-afternoon. The outage hit a datacentre in Google’s europe-west2-b zone — one of more than a dozen zones Google maintains in Europe to service its clients. Because Google Cloud stores data and runs operations for a number of businesses, the downtime produced widespread knock-on effects that were mostly confined to UK businesses.

Multiple banking services faltered, court hearings were put on hold and patients reported temporary glitches on the NHS app, according to user submissions tracked by Downdetector. Royal Mail, Revolut, HSBC and Lloyds bank were among the services reported as affected. Revolut confirmed “a service disruption affecting multiple services” but added that “all our customers’ money and data are safe.” Lloyds said its app was working but acknowledged that “some [people] are having issues with certain online services this morning.”

HM Courts and Tribunal Services, which handles remote video hearings in courts across England and Wales, said it was “aware of an issue affecting Cloud Video Platform” and was “working with our technology partners to restore service.”

The Guardian has reported that up to 60% of the UK government’s infrastructure is hosted on cloud services, mostly with Amazon, Microsoft and Google. British police keep vital data on Microsoft’s Azure cloud, the paper has reported, despite being apparently aware that this exposes sensitive information to US government insiders, hostile foreign actors and outages like Tuesday’s.

The incident echoes a far larger outage roughly a year earlier, when Amazon Web Services’ us-east1 region in Virginia — the densest concentration of datacentres on Earth — went down for about 15 hours. That cascade took out the Roblox game and the Signal messaging app and caused internet-connected mattresses worldwide to heat overnight and toast their occupants.

Last year’s AWS outage was later revealed to have been caused not by a power failure but by a bug in automation software. A separate AWS outage was traced to an AI tool that Amazon had adopted after laying off tens of thousands of staff the previous autumn.

Internet infrastructure expert Michał ‘rysiek’ Woźniak said the pattern shows hyperscalers had become “too big to not fail.” “But worse,” he added, “they’ve become so deeply ingrained into the very fabric of our digital lives that switching away from them seems unimaginable.”

Woźniak pointed to a Google announcement from earlier this year indicating that 75% of all new code at the company is “now AI-generated and approved by engineers.” “One can only wonder how that affects the reliability and quality of Google’s services,” he said. “Multiple large outages at another hyperscaler, AWS, a year ago have been linked to internal AI tooling.”

Woźniak also said: “I am sure Google’s engineers are doing everything they can to mitigate them, but it is unreasonable to have a handful of gigantic companies whose outages time and again cause public institutions or large swaths of the digital economy to stop working en masse.”

Google has been approached for comment.