
Google's object storage: multi-region, automatic classes with Autoclass and undelete — the lake to build data and AI on.
Cloud Storage is object storage with Google's finesse: multi-region buckets with transparent failover, Autoclass moving objects across classes (standard→nearline→coldline→archive) based on real use, soft delete and versioning for the incidents, and the bandwidth of Google's network to serve the world.


Buckets, prefixes and uniform permissions (IAM, not ACLs): the lake that doesn't become a swamp.
The right class per object without rules to write: the intelligent default.
Versioning, soft delete and retention locks: the bucket that forgives (and resists ransomware).
BigLake and external tables: the storage analytics queries without copies.
Cloud Storage has four classes (Standard, Nearline, Coldline, Archive) under a single constant-latency API: Autoclass moves objects on its own based on access; versioning and soft delete recover mistakes, the retention policy with bucket lock (WORM) armors backups and compliance, dual/multi-region replicates for DR with 15-minute turbo replication. Uniform IAM and signed URLs govern access.
The base of BigQuery, Spark and Vertex AI.
The immutable copy outside the perimeter.
Media and downloads served by Google's network.