Google Cloud

The third hyperscaler, from Google’s own geography-and-regions documentation (T3 — first-party infrastructure description).

The vocabulary is the same shape as Azure’s and the guarantees are stated more crisply. Regions are “independent geographic areas that consist of zones”; a zone is “a deployment area for Google Cloud resources within a region” and “should be considered a single failure domain within a region.” A region normally contains “three or more zones housed in three or more physical data centers” — with named exceptions still building out (Stockholm, Mexico, Osaka, Montreal). Coverage spans North and South America, Europe, Asia, the Middle East and Australia.

Some services are multi-region by construction — BigQuery, Cloud Storage, Firestore, Spanner — and are “designed to be able to function following the loss of a single region.”

The operational advice is the useful part, because it is the vendor telling you its own failure model: deploy across multiple zones for availability, and for disaster recovery either use a managed multi-region service or run your own replication to another region.

Reading the three together

With aws, azure and this page, the spoke can finally state what a hyperscaler sells that a fixed-price VPS does not, without generalizing from one example:

• a failure domain hierarchy made explicit and contractual (zone → region → geography),
• managed multi-region state — the hard part, and the real lock-in, since Hetzner will rent a second machine but not replicate your database across continents,
• a residency boundary you can name to a regulator.

Google’s “three or more zones in three or more physical data centers” is the most specific commitment of the three, and it names the regions that do not yet meet it.

Not read here: Google Cloud pricing or the sustained-use/committed-use discount model, which is where the per-hour comparison against the mid-market actually happens.