# Google — `gcp-regions`, `google-datacenters`, `google-news`

## gcp-regions (Google Cloud regions)

| | |
|---|---|
| Source | `https://cloud.google.com/compute/docs/regions-zones` (static docs table `Zones | Location | Machine types | …`) |
| Kind / mode | `cloud_provider` / `html`, implementation `cloud_regions`, parser `cloud_gcp_regions_v1` |
| Fetch | L1 direct, 0 credits (cloud.google.com robots.txt allows /compute/docs) |
| Schedule | weekly |
| License / attribution | Provider documentation (Google developer docs are CC BY 4.0); facts only — "Google Cloud" |
| Verified live | 2026-09-11: 130 zones → 43 regions, 43 valid, 0 rejected |

Zones (`europe-west1-b`) are grouped on their prefix → region code; the zone count is `availabilityZones`
(us-central1 = 4, everything else 3); the `Location` column ("St. Ghislain, Belgium, Europe") gives city + country/state.
Spot checks: `europe-west1` → St. Ghislain BE, 3 AZs; `asia-east1` → Changhua County TW (precision `approximate`,
a county); `northamerica-northeast1` → Montréal, Québec CA; `us-east4` → Ashburn, Virginia US.

**Gap.** Only live regions appear in the docs table. Announced GCP regions are not published statically:
`/about/locations` is a JS shell over `data.json` (service × location, no geography, no "coming soon") — deliberately
not crawled. `launchedOn` is not published.

## google-datacenters (Google-owned facilities)

| | |
|---|---|
| Source | `https://www.google.com/about/datacenters/locations/` — list embedded server-side in `<div id="data-centers-data">` |
| Kind | `cloud_provider`, kind `facility`, parser `cloud_google_datacenters_v1` (operator Google, `hyperscale`, `isHyperscale`) |
| Fetch | L1 direct, 0 credits (robots.txt: /about/datacenters allowed) |
| Verified live | 2026-09-11: 67 sites, 67 valid, 0 rejected (36 operational, 31 "in development" → `announced`) |

Each object: `location` ("Council Bluffs, Iowa"), `inDevelopment`, sometimes `latitude/longitude`, `detailPage`.
Status: Google's own "in development" label → `announced` (the page does not say whether construction started);
otherwise `operational`. Geo: curated city table for towns (46, precision `city`); counties / states use the curated
county marker or Google's own coordinates, flagged `approximate` (21). Same-name towns are guarded by the state
("Lebanon", "Columbus", "Douglas County" only resolve when the state matches). Detail pages 404 without the exact
path and are not fetched. Names are "Google <location> data center"; keys `google:dc:<slug>`.

## google-news (The Keyword)

Feeds `blog.google/inside-google/infrastructure/rss/` (precise: "€13 billion investment in AI infrastructure in
Finland", Alabama/Virginia investments, energy deals) and `blog.google/products/google-cloud/rss/` (filtered on
title/summary before fetch; consumer products excluded). Live 2026-09-11: 11/40 items matched, 5 entities incl. one
project candidate, 0 rejected, 0 credits. Daily.