One power event in a single London cloud zone: how Revolut, HSBC and Lloyds went down

On Tuesday, October 20, 2026, a power failure in the europe-west2-b zone of Google Cloud's London region disrupted services across the UK. Google confirmed the root cause was a "power event" and said engineering teams restored full service availability. Services began returning around 1pm London time and were back online for most customers by 3pm.

The downstream list is the story. Banking apps at Revolut, HSBC, and Lloyds left customers unable to access or send cash. The remote video system used for court hearings across England and Wales stalled. A property listings site, a parcel courier, and UK public-sector websites were also affected.

Revolut confirmed a disruption that blocked card payments, ATM withdrawals, and app sign-in. A person familiar with the matter said Revolut's outage was tied directly to the Google Cloud disruption. Revolut said customer money and data remained safe. Its 3.4 million Irish customers were affected, with card payments, transfers, cash withdrawals, and account sign-in all hit.

Only Revolut's link to Google Cloud is attributed to a person familiar with the matter. HSBC and Lloyds have not officially confirmed the cause, and much of the early picture came from Downdetector user reports rather than official confirmations from every company involved. What caused the power event in europe-west2-b, and what redundancy failed, is not yet public.

The incident renewed regulatory concern in the UK and EU about the financial sector's dependence on a small number of external technology providers, particularly cloud firms supporting critical banking services. That concern is well founded. When one hyperscale zone misbehaves, the damage does not stay on the provider's status page. It shows up as a declined card at checkout, an app that will not open, or a stalled court hearing.

The gap between Google's narrow "power event, now mitigated" language and the breadth of reported disruption is the practical lesson. A single-zone failure propagated into civic and financial infrastructure that most users never associate with a cloud region.

• Map every critical service to the specific cloud zones and regions it depends on, and identify single-zone dependencies in payment, identity, and communications paths.
• Test failover across zones and regions, not just within a region. Confirm that DNS, load balancing, and database replication actually shift traffic when a zone drops.
• Review third-party and fourth-party concentration. If multiple vendors share one hyperscaler region, treat them as one dependency.
• Build degraded-mode procedures for customer-facing teams: manual card authorization fallbacks, offline court scheduling, and status-page messaging that reflects downstream impact, not just provider health.
• Monitor provider status feeds alongside independent signals such as Downdetector and internal error rates, and reconcile the two before communicating.
• Rehearse incident communications that state what is unknown. Overclaiming a narrow root cause while customers cannot pay is its own outage.