Google Cloud’s London availability zone europe-west-2-b lost power for 7 hours and 52 minutes on October 20, 2026, after an underground utility cable failed and a programmable logic controller misconfiguration stopped the emergency generators from starting, knocking out banking apps, courier services and court video links at Lloyds, HSBC, Revolut and Royal Mail.
“Circuit breaker feeder B was physically opened to allow for corrective repairs following issues identified during a quarterly generator load test on 14 Oct 2026,” Google admitted. The admission, published in a November 4 update to its service status page and reported by The Register, sits at the center of the most disruptive London cloud outage of the year.
Why the Generators Never Started in europe-west-2-b
The facility runs a 2N utility power architecture — complete duplicate power paths for every critical load — fed by two independent utility connections Google calls “A” and “B,” backed by a central high-voltage generator bank managed by a dual-redundant PLC. On October 20, an external underground cable carrying power to feeder A broke at the exact moment feeder B was unavailable.
That should have been survivable. The zone carries uninterruptible power supplies capable of holding the critical load for five minutes, and on-site generators exist precisely for this scenario. The UPS units worked. The generators did not.
“PLC logic indicated that a healthy utility source was available, even though the primary and secondary distribution feeds were de-energized, and did not issue a start command to the generator plant,” Google’s status update revealed. A voltage-sensing design flaw compounded the misconfiguration, so the five minutes of UPS runtime expired with the generator plant still silent. The rest, as The Register put it, is history.
Feeder a Broke During a Planned Repair Window on Feeder B
The timing is what turns a routine cable fault into a seven-hour outage. Feeder B’s circuit breaker had been physically opened six days earlier, during the quarterly generator load test on October 14, to allow corrective repairs on the B feeder pathway. With feeder A carrying the entire critical load and feeder B deliberately offline, a single cable failure removed all utility power at once.
Google has since reprogrammed the PLC so it cannot misread the same conditions, reconnected to its power utility, and says it has tested the changes at europe-west-2-b and confirmed they work. The disruption itself ran from 7:08 a.m. to 3 p.m. British Summer Time, with errors and network failures logged from 06:08 UTC, and touched 38 products including Compute Engine, Google Kubernetes Engine, Cloud SQL and Cloud Storage. Google advised customers to move workloads elsewhere where they could.
One question remains open: why the region’s two healthy availability zones did not fully absorb the load, a point a developer write-up citing Computer Weekly raised and that Google’s pending analysis is expected to settle, windowsforum.com reported.
Banks Can Outsource Infrastructure but Not Resilience
The customer list reads like a cross-section of British daily life: Revolut, with roughly 13 million UK customers, saw logins, card payments and cash withdrawals fail, though it said customer funds and data stayed safe; Lloyds, HSBC and Halifax acknowledged access problems; Royal Mail reported problems while insisting letter and parcel deliveries were unaffected; Rightmove, DPD, NHS England and court video links also spiked on Downdetector, which windowsforum.com cautioned aggregates user reports and may overstate actual impact.
The outage has reignited the cloud concentration debate. Raphael Auphan, chief operating officer at Proton, put the dependency bluntly:
“Europe is almost entirely dependent on a handful of cloud companies.”
He argues for more investment in regional providers.
Jamie Beckland of APIContext argued that banks can outsource infrastructure but not responsibility for resilience. Graeme Stewart of Check Point compared the digital economy to millions of organisations plugged into one extension lead. The argument is persuasive, but it has limits. Diversifying across providers does not remove the risk that an organisation’s own architecture cannot actually fail over. Calling for multi-cloud is easy; a tested, working recovery path is harder to build.
The PLC Fix and the Two Action Items Google Set Itself
Google has committed to two concrete follow-ups. First, a “Communications Enhancement”: an automated alert-to-communication mechanism for location-impacting incidents, so customers hear about troubles faster. Second, expanded platform-wide testing outcomes for infrastructure-level events — in plain terms, rehearsing the failure modes this incident exposed rather than discovering them live.
For anyone running workloads on Google Cloud, Azure or AWS, the practical lesson from October 20 is specific: even a 2N power design with dual-redundant controllers can fail when a maintenance window on one feed coincides with a hardware fault on the other, and when the control logic misreads a dead grid as a healthy one. The next dated milestone on record is Google’s promised completion of those two action items — and its full incident analysis explaining why the healthy zones didn’t catch the load.
