Google Cloud: a power and cooling failure hits a europe-west4 data center for 15 hours

2026-07-21 · Publisher : Google Cloud · Domain : Cloud & DevOps

In the week of 13 July 2026, an electrical fault upstream of a Dutch data center makes the backup power fail, then the cooling. The room reaches 44 °C; three services are unavailable for about 15 hours.

### Background In the week of 13 July 2026, the data center hosting three services (VMware Engine, Bare Metal Solution and NetApp Volumes) in **Google Cloud** zone **europe-west4-a**, in the Netherlands, suffered an outage of about **15 hours**; the rest of the zone kept working. According to The Register, drawing on Google's incident report (article of 21 July), an electrical fault on the utility grid upstream disrupted power distribution. The side A backup system, a rotary uninterruptible power supply (DRUPS), did not take over the load because of "electrical component failures". The chiller controller also went offline during the voltage transient. ### Key facts - Backup cooling was unavailable because of ongoing construction work: the room temperature reached **44 °C**, and machines were shut down as an emergency measure. - Services affected: Google Cloud VMware Engine, Bare Metal Solution and NetApp Volumes. - Google quote: "We know how much you rely on Google Cloud, and we regret the impact on your productivity." A final report is to detail preventive actions. - The Register's argument: hyperscalers are not transparent about single-data-center dependencies inside a zone, hence a false sense of security for managed services. ### Why it matters As with the AWS outage in May, it was physical infrastructure that failed, not software. The services concerned (VMware, bare metal, storage) are typically migrated as-is, without re-architecting, which can tie them to a single zone. For IT leaders, two questions for providers: does a managed service depend on a single site in the zone, and which site works can neutralize the advertised redundancy. ### Takeaway Advertised power and cooling redundancy did not hold: single-zone services remain exposed to the loss of a site.

Tags : Incident, Google Cloud, Datacenter, Hyperscaler