Google Cloud’s London outage was six hours long. Your second site should not be in the same cloud

A power failure took a Google Cloud London zone down for six hours. Why critical services need an independent bare-metal site, plus five checks.

Short answer: On October 20, 2026, one availability zone of Google Cloud’s London region lost power for about six hours. A design flaw and a controller misconfiguration kept the generators from starting, the batteries ran out and 38 products were affected across the London region. Google published its incident report on November 4. The lesson is not to leave the cloud. It is to keep critical services on an independent failure domain, such as bare-metal servers in a second facility you control.

As Capacity reported on November 13, the effects reached daily life: shops that could not take card payments, disrupted parcel deliveries and court hearings postponed when video links failed. Samuel Carvalho, CEO of TelCables Europe, summed it up: “Failures are inevitable. Outages of this scale are not.”

One zone, one provider, one grid connection

Power failures happen in every data center. The damage grows when payments, logistics and communications depend on the same zone, the same provider and the same utility feed. Spreading workloads across zones inside one cloud helps, but those zones still share a provider, a control plane and often a city.

Where dedicated hosting fits

Bare-metal servers in a separate facility give you a second failure domain that does not depend on your primary cloud. Keep the critical path there: payment authorization, DNS, messaging gateways, a read replica of your core database or a warm standby of your application.

What US Matrix Dedicated Hosting provides

  • Single-tenant bare-metal servers in the country you choose, in 30+ countries.
  • Carrier-neutral partner data centers, independent of your cloud provider.
  • Out-of-band console and remote power control that work even when the network is down.
  • Redundant switching, dedicated internet access, 1 and 10 Gbps uplinks, IPv4 and IPv6.
  • 24x7 monitoring and NOC, unmanaged or managed.
  • One contract, one invoice and one support team for every location.

Five checks for your resilience plan

  1. List the services that stop your business if one cloud zone goes dark.
  2. Check whether your backup runs on a different provider, site and power feed.
  3. Test a failover to the secondary site at least once a quarter.
  4. Make sure you can reach your servers out of band when the network fails.
  5. Agree in writing who answers, and how fast, during an incident.

Resilience is not a second zone. It is a second answer when the first one goes dark.

Tell us where you need servers: usmatrix.com/contact.

Qualquer um lança uma loja de eSIM em minutos. Veja onde está a margemA FCC acaba de desligar oito provedores de voz. Algum deles está na sua rota?Anyone can launch an eSIM store in minutes now. Here is where the margin isThe FCC just cut off eight voice providers. Is one of them inside your route?Google Cloud’s London outage was six hours long. Your second site should not be in the same cloudA queda do Google Cloud em Londres durou seis horas. Seu segundo site não deveria estar na mesma nuvemA vacância de data centers caiu quase pela metade. Garanta seus racks antes da IAData center vacancy nearly halved this year. Secure your racks before AI does