Three major cloud platforms have shipped Postgres-flavored databases in the past year, each promising to bridge operational and analytical workloads. Snowflake Postgres is now generally available, built on Crunchy Data's work with pg_lake for lakehouse integration. Databricks Lakebase launched on AWS with a Neon-derived branching model enabling instant database branches for CI/CD. Azure HorizonDB remains in invite-only preview, with Microsoft claiming up to 3,072 vCores and 128 TB databases.

All three claim to deliver " Postgres for the AI era" but wire-compatible isn't the same as actually Postgres. Snowflake offers the most recognizable Postgres engine with decent extension support. Lakebase's branching model is genuinely useful for developers needing point-in-time recovery as a normal operation. HorizonDB is architecturally ambitious—Microsoft built their own storage engine from scratch—but the gap between wire-compatible and actually Postgres grows with your dependency on extensions and tooling.

The honest recommendation: pick the platform adjacent to your existing data infrastructure. If your warehouse runs Snowflake, take Snowflake Postgres. Databricks shops get Lakebase. Azure users tired of VM management consider HorizonDB. What you gain is operational scale and tighter integration; what you lose includes flexible extension support, predictable upgrade timing, and tools like pg_basebackup that don't apply to shared-storage architectures.

Most production Postgres workloads still fit comfortably on a single primary with replicas. The shared-storage scale-out story is real, but it's a real story for a small fraction of workloads. Don't bet your operational stack on preview software.