What Happens to Six GPU Cloud Providers Under 72-Hour Sustained Load

We ran identical inference workloads across six providers for three days straight. Two throttled. One crashed. Three held.

(Full lab note to be written — this is a migrated stub so the route and listing work.)