Download websites that use Google Cloud

A site can appear on this list without anyone at the company having opened a cloud console. The workload might belong to an ecommerce platform, an agency hosting product, a marketing automation vendor or an embedded widget, any of which can put a customer domain onto infrastructure the customer never chose and could not name. A hosting detection is a statement about where bytes are served from, and in a world of layered software services the serving party and the buying party are frequently not the same organisation.

Accepting that makes the row more useful rather than less, because the delivery fact is exactly what infrastructure adjacent vendors sell against. Observability, cost management, security posture, data transfer and migration are bought by whoever operates the workload, and the operator is who the fingerprint identifies. Offered by Google since 2008 and sold as Proprietary software, the platform has a service catalogue broad enough that a single positive detection can mean a virtual machine estate, a managed container runtime or one serverless endpoint behind a load balancer.

Vendor consolidation is where the row earns its keep. Our crawl records mail routing, content delivery and analytics on the same domain, so you can separate domains sitting inside one supplier's estate at several layers from those that split their stack. A company that serves content, routes mail and runs compute with a single supplier is a different prospect, with a different switching cost and a different procurement path, from one that mixes.

As of November 2026 we detect the platform on 4,994,523 sites, 2,697,816 of them in United States, with the set skewing toward .com. Concentration of that kind follows region availability and the incorporation of the software layer sitting on top, since a hosted application serving customers worldwide still registers as one country when its own domain is measured.

Cloud attribution is harder than it looks and we would rather be plain about it. A delivery network or reverse proxy in front of an origin hides the origin entirely. Managed and serverless products often leave no distinguishing response signature. Estates spread across several providers produce a partial picture, because the crawler sees whichever component answers the request for the hostname it asked about and nothing else. Published address ranges also change hands over time. Read the row as evidence of presence, never as an estimate of spend or of share.

One domain can be resolved and inspected in a minute, so the value of a downloadable list is never the single answer. The value is the filtered population, where the infrastructure signal is combined with country, suffix, platform and the other detections to produce a target set worth working, and where comparing snapshots turns movement between providers into rows that changed rather than into a rumour someone heard at a conference. For a sales team the output is a territory. For a researcher it is a measure of how much of the public web depends on how few suppliers, which is a question no per domain lookup can approach and which becomes more pressing every year.