Google Cloud Says a Voltage Drop in the Utility's Feed, and the Breakers' Protective Action, Caused a Service Outage in One Zone

On July 25, 2026 Google Cloud posted its incident report on an outage in its europe-west4-a zone, in the region its Service Health page lists under the Netherlands. The report names the root cause as "a 3ms voltage drop in the power feed from the utility provider and subsequent utility breaker protective action." Customers experienced disruptions across several of the company's products, and the interruptions lasted 14 hours and 55 minutes in total.

The summary says the data center provider rectified the failed systems, and Google's engineering team then restored the services. In a preliminary report posted on July 17, Google's engineers worked with "the third-party facility provider" to restore primary utility power and cooling. Parts of the incident page follow, not the whole.

The Dip and the Backup

The data center, the report says, "experienced an upstream voltage transient affecting both utility power feeds A and B." Its timeline of data center events, kept in Pacific time, opens at 16:24 on July 15 with an electrical fault on the utility grid upstream of the site. The utility breakers on both feeds tripped and started the transfer to the backup source, a DRUPS, which the report spells out as a "Diesel Rotary Uninterruptible Power Supply."

Side B made that move "without any power interruption." Side A did not. Its DRUPS, the report says, "failed to take over the facility load due to electrical component failures."

That failure set off an automatic transfer of the three affected rows from feed A to feed B. Rows 1 and 2 moved across. Row 3 did not, and lost all power when an overload protection breaker tripped. The report gives the root cause of that failed transfer as "load deployment discrepancy which is under further investigation by Google engineering team," and does not explain the phrase. The row was left with neither of its two redundant feeds.

Two feeds protect a row only as long as the move from one to the other works. For one row in this hall, it did not.

No Water Flow

The cooling failed in the same event. The chiller controller dropped offline during the transient, so the chilled water distribution pumps got no signal to restart, and chiller system A shut down. The redundant source, the report says, "was not available due to known ongoing construction work at the facility."

The hall heated. At 18:29 in the same timeline, data hall temperatures "reached 44C and crossed the safe machine operating threshold."

Under its heading "Remediation and Prevention," the report comes back to the chillers. They "maintained power but shut down because the distribution pumps failed to restart, resulting in no water flow." The chillers had power. The water did not move. The pumps were switched manually from auto to hand mode to get the water circulating again, and the facility team deployed "an interim portable UPS to support the chiller controllers." The data center timeline has the cooling system fully recovered at 21:05.

Turning the Machines Off

As the hall heated, Google's own teams shut equipment down, beginning machine shutdown procedures "for the remainder of reachable devices as part of the cooling emergency shutdown process." The updates the company posted while the outage ran say it "proactively turned down workloads in order to protect customer data from any risks posed by running infrastructure in a high temperature environment."

Those updates also told customers that no workaround was available. Customers "with multi-regional deployments," they said, were advised to route traffic to an alternate site. In the report's description of impact, the disruptions ran "from 16:39 to 16 July 2026 07:34 US/Pacific."

Remediation and Prevention

On the power side, the facility team aligned "a swing DRUPS unit to restore the redundant configuration on the A feed" and is working on repairing the faulty one, the report says. Google then lists the actions it "is completing," each with an ETA.

An investigation of the sequence of events that prevented the transfer of load to the DRUPS carries August 2026. So does the work to find and resolve "the cause of overload conditions that caused power loss to Row 3." An investigation of the redundancy setup in the chiller pump controls carries September. Work by one service's engineering team to improve the automated triggering of shutdown "during quickly evolving thermal runoff situations" carries October 2026. The months are the report's ETAs, as it gave them on July 25.

What will the investigation of the failed transfer on Row 3 find?