diff --git a/CHANGELOG.md b/CHANGELOG.md
index cf9a083bd1..c4f04f8c12 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -26,6 +26,16 @@ aggregate instead: an italic *Catalog* line at the end of the version section an
 
 ## [Unreleased]
 
+### Changed
+
+- **`anyplot-app` scales to zero** — the frontend service ran a permanently warm
+  instance for ~EUR 8.30/month while 99.56% of the paid time was idle. It is a static
+  nginx image that boots in ~0.26 s, and a 7-day request trace at one-minute resolution
+  shows the longest gap between requests is 11 minutes, against Cloud Run's ~15-minute
+  idle window — so the instance is in practice never reclaimed and visitors keep the
+  same time to first byte. `anyplot-api` keeps `min-instances=1`: its cold start is
+  ~11.6 s and its traffic does leave gaps over 15 minutes. (#10812)
+
 ## [3.2.0] — 2026-08-29 — Findable by assistants
 
 anyplot 3.2 makes the catalogue findable by assistants, not only legible to them. The sister
diff --git a/app/cloudbuild.yaml b/app/cloudbuild.yaml
index d749fb7219..83914a107f 100644
--- a/app/cloudbuild.yaml
+++ b/app/cloudbuild.yaml
@@ -53,8 +53,13 @@ steps:
       - '1'
       - '--timeout'
       - '60'
+      # Scale to zero: this is a static nginx that boots in a fraction of a
+      # second, and its traffic is dense enough that Cloud Run rarely reclaims
+      # the instance anyway, so a warm one bought almost nothing. The API keeps
+      # min-instances=1 — its cold start is orders of magnitude larger.
+      # Measurements behind this: #10812.
       - '--min-instances'
-      - '1'
+      - '0'
       - '--max-instances'
       - '3'
       - '--port'