A power failure in europe-west2-b, one of three zones inside Google Cloud's only UK region, took down banking apps, the courts' video platform and delivery networks on 20 October 2026. With no second UK region to fail into, the incident should prompt every operator to revisit their downtime cost calculator assumptions.
View the data behind this chart
| Single-region cloud | Multi-region cloud | Hybrid/on-prem DR | |
|---|---|---|---|
| Zone outage failover | No failover available | Possible if app multi-region | Independent of cloud zone |
| Recovery time | Hours (as seen 20 Oct) | Minutes if tested | Minutes to near-zero |
| Provider dependency | Single vendor and region | Single vendor, multi-region | Vendor-independent |
| Data sovereignty | UK-only, no backup zone | UK plus EU failover option | On-prem UK, full control |
| Typical cost | Lowest cloud spend | Higher egress/compute cost | Capex plus maintenance cost |
What happened in Google Cloud's London region
Just after 07:05 BST on 20 October 2026, the europe-west2-b availability zone inside Google Cloud's London region lost power. According to engineers tracking the incident, backup generators failed to engage, and the two peer zones, europe-west2-a and europe-west2-c, did not fully absorb the displaced workload, leaving Apigee-dependent services among the hardest hit.
Google confirmed a power event in the zone and said it was mitigated by 12:27, with partial restoration following within 30 minutes. By mid-afternoon Google reported recovery for 'most workloads' in europe-west2, though it described the customer impact as a 'very small number' of accounts. The Ministry of Justice confirmed at 15:52 that its Cloud Video Platform had still not fully recovered.
Why a single UK region is a single point of failure
The structural issue is straightforward: europe-west2 is Google Cloud's only UK region. Its three zones give resilience against a single rack or room failing, but they all sit inside the same metro footprint and, as this incident showed, can share the same cause. When a zone goes down and its peers can't absorb the load, there is no second UK region to route around — only EU options, which reopens data residency and regulatory questions for banks, courts and health bodies that chose UK cloud specifically for sovereignty reasons.
That trade-off is worth running through a cloud vs. on-premise TCO calculator before the next renewal. Many UK buyers assumed zonal redundancy equated to regional resilience; this outage shows the gap between the two, and it is a gap that sits squarely in the architecture, not in any single vendor's uptime SLA.
Who felt it: banks, courts and the supply chain
The outage rippled across sectors that rarely share an incident report. Banking and fintech names affected included Revolut, Lloyds Bank, HSBC, Barclays, Bank of Scotland, NatWest, Halifax, Nationwide, Emma and credit-reference firm Equifax. Delivery and retail operators Royal Mail, DPD and Screwfix were hit alongside Rightmove, Cloudflare, BT and hosting firm SiteGround. In the public sector, NHS England, HM Courts & Tribunals Service and Gov.uk all showed disruption.
The courts' Cloud Video Platform went down across large parts of England and Wales from 10:00, with the Criminal Bar Association warning that hearings could not proceed. This lands against a backdrop where investment banks alone are estimated to lose £600,000 an hour to IT failures — a figure that makes the case for running your own numbers through a downtime cost calculator rather than relying on a provider's headline uptime claim.
Multi-cloud and on-prem failover: validating your DR plan
For regulated and customer-facing services, the practical response is to stop treating a single cloud region as a DR strategy. That means testing failover to a genuinely separate region or provider, and for the most critical workloads, keeping a hybrid fallback that doesn't depend on the same control plane. Reviewing hot, warm, and cold DR sites options against your recovery time objectives is a sensible starting point, as is comparing DRaaS vs. self-managed DR for the workloads that genuinely cannot tolerate hours of downtime.
On-premise or colocation capacity, backed by third-party maintenance services or cost-effective refurbished servers, remains a credible fallback tier for firms unwilling to bet an entire service on one cloud region. Running the exercise through a backup and DR sizing calculator turns this from a theoretical debate into a budgeted plan.
A repeating pattern: vendor concentration risk
This is not an isolated event. It echoes an AWS outage last autumn that knocked out UK tax office and banking services, and it reinforces warnings that the UK's largest businesses remain dangerously dependent on a handful of cloud providers. As one administrator following the Google incident put it, 'It is the only Google region in the UK so businesses/banks/NHS etc. here can't switch to another location.' That single sentence captures the architectural reality every UK buyer now needs to plan around.
What UK infrastructure buyers should do now
The immediate actions for IT and risk teams are operational, not theoretical.
- •Map every critical service against its actual region and zone dependency — not just the provider name.
- •Test failover to a second region or provider at least annually, including for third-party SaaS dependencies like Apigee-style gateways.
- •Reassess data sovereignty trade-offs where the only 'UK' option is a single region with no in-country backup.
- •Quantify exposure through a downtime cost calculator and size DR spend accordingly, including backup tooling such as Veeam.
- 01Computer Weekly — Google Cloud power event knocks out UK banks, courts and couriers · 20 October 2026
- 02ITPro — The Google Cloud outage explained · 20 October 2026