On the morning of 20 October, a lot of people tapped their phones at a till and watched nothing happen. A power failure at a London data centre run by Google Cloud knocked out Revolut, which has around 13 million UK customers, along with several high-street banking apps, Royal Mail tracking and DPD’s delivery service. Nobody’s money vanished. It just became unreachable for a few hours, which, if you’re standing at a checkout with a queue behind you, feels identical.
The odd part is that none of those companies lost a single server of their own. They rent computing power from someone else, and someone else had a bad morning. That chain of borrowed infrastructure is the real story.
What “the cloud” physically is
The cloud has a sort of fluffy reputation, but it is warehouses. Rows of servers, cooling plant, backup batteries, diesel generators, fibre cables. When a banking app loads your balance, a request travels to one of those buildings, a machine does the sum and the answer comes back in a fraction of a second.
The building can fail. A power event, a cooling fault, a flooded basement, even a drone strike: Amazon’s UAE sites were hit earlier this year, and banks and payment apps in the region wobbled straight away. Software bugs get the headlines, yet plain physical damage is just as capable of freezing a payment.
London made the October failure worse. As reported at the time, the capital had a single Google Cloud region, so companies that wanted to fail over had nowhere local to go. Moving traffic to a data centre in another country is possible, but it brings data-residency rules and latency headaches that most firms hadn’t rehearsed.
Why it keeps happening
This was not the first time. The anniversary of the 2025 AWS outage fell on the very same day, and that one had Lloyds, Halifax and Bank of Scotland customers locked out of their accounts. Before that came the 2024 CrowdStrike fiasco, which had little to do with a data centre and everything to do with one faulty update spreading through bank branches and call centres at once.
The pattern is concentration. A handful of providers host a huge slice of the internet, so a fault at one of them travels far. The Bank of England has warned regulators that this reliance could threaten financial stability, because the same few providers sit underneath so many rival firms. Competition on the surface, shared plumbing underneath.
Switching banks does not fix this. Your old bank and your new one may run on the very same cloud region.
What uptime really means
Every service boasts about uptime, and the numbers look impressive. 99.9% sounds basically perfect. Do the sum, though: 99.9% allows roughly 8.7 hours of downtime a year. 99.99% allows about 52 minutes. 99.999%, the “five nines” figure that engineers chase, is around five minutes.
Three catches. First, uptime is usually averaged across a whole year, so one nasty four-hour morning can sit comfortably inside a generous promise. Second, it measures whether the service responds, not whether it responds well. A banking app that loads but cannot show your balance counts as “up” in plenty of dashboards. Third, the guarantee often applies to the provider’s own infrastructure, not to the apps built on top of it, so a firm can quote five nines while depending on a supplier that quotes three.
Redundancy is the real answer to all this. Run the same service in two buildings, ideally in different regions, and switch automatically when one fails. It costs real money, because you’re paying for capacity that mostly sits idle, so smaller firms often skip it and hope. Most days, hope works.
How to judge reliability from the outside
You can’t inspect a server room, but you can read the signals. A decent checklist for any online service that handles your money:
- Published status history. Firms that keep a public incident log, with dates and causes, are telling you they expect scrutiny. A blank history is not a clean one.
- How they talk during an outage. Fast, specific updates beat vague apologies. Silence for three hours says something.
- More than one way to pay or withdraw. A second payment rail turns a total blockage into an inconvenience.
- A licence you can verify. Regulated firms must report major incidents and hold customer funds in a defined way.
- A human on the other end. Live chat or a phone line that still works when the app doesn’t.
Apply that list to any category and the good operators separate quickly from the rest. That includes online entertainment aimed at UK adults. A bingo site is, after all, just another app that takes a deposit, holds a balance and pays out later, with everything that implies about uptime. The comparison pages that rate them on licensing, payment options and support response are doing the same reliability homework: https://europeangaming.eu/portal/best-bingo-sites-uk/ lists UK bingo sites against exactly those practical criteria, so a reader can see which ones offer several payment methods and a working support channel before depositing. Gambling involves risk, so play responsibly and only with money you can afford to lose.
What you can do before the next outage
Plan for the dull failure, not the dramatic one. Practical habits that cost nothing:
- Keep a second payment method that does not share a provider with your first. A physical card from a different bank beats three apps from the same one.
- Carry a small amount of cash. It feels old-fashioned, and it works when the network doesn’t.
- Know your bank’s digital money tools well enough to find offline options, such as saved payee details or a printed account number.
- Screenshot important confirmations. When a service recovers, records occasionally lag behind.
- Pay bills a few days early. A delayed payment during an outage is annoying; a missed deadline is worse.
None of that prevents an outage. It just stops one from becoming a bad day.
What the October failure should change
Companies will say the right things. Post-incident reports will promise more regions, more testing and better communication. Some of that will be real. The cynical read is that redundancy gets funded right after a failure and trimmed again two budgets later.
For consumers, the sensible takeaway is modest. Cloud services are mostly very reliable, and they fail in ways that are rare, short and fixable. But “mostly” is carrying a lot of weight. If one data-centre fault can freeze a neobank, three high-street banks and a delivery network on the same morning, then the idea that your money is always one tap away deserves a footnote.
FAQ
What caused the October 2026 UK outage? A power failure at a Google Cloud data centre in London disrupted services hosted there. Revolut, several banking apps, Royal Mail and DPD were affected. The fault was physical rather than a cyberattack, and services returned once power and systems were restored.
Is my money safe during a cloud outage? Yes, balances are not erased. They sit in records that become temporarily unreachable. Regulated firms must protect customer funds separately, so an outage blocks access rather than ownership. Once systems recover, balances and pending transactions normally reconcile.
What does 99.9% uptime actually mean? It allows about 8.7 hours of downtime per year. Higher figures shrink that: 99.99% allows around 52 minutes. Check how the provider measures it, since a service can be technically up while key features still fail.
What is redundancy in cloud computing? Running the same service in more than one data centre, ideally in separate regions, so traffic can switch if one fails. It is the main defence against outages, though it costs more, which is why not every company does it.
How can I protect myself from payment app outages? Keep a second payment method from a different provider, carry some cash, pay important bills early and save confirmation screenshots. These habits cost almost nothing and turn a total blockage into a minor delay.
