Snowflake

The AI data cloud

Cloud data warehouse with separate compute and storage, zero-copy cloning, and Data Cloud ecosystem.

Our take

Snowflake is the cloud data warehouse that defined the modern data stack, and its pricing model — separate compute and storage, instant elasticity, pay-per-query — is still competitive a decade in. Zero-copy cloning lets you create development and test environments from production data in seconds without duplication cost. Time Travel (point-in-time queries up to 90 days) is a practical disaster recovery and audit tool. Snowflake's Data Cloud ecosystem — Data Marketplace (pre-built datasets), data sharing (zero-copy cross-account sharing), and Snowpark (Python/Java/Scala in Snowflake) — extends the platform beyond query engine. The cost critique is real: Snowflake credits burn fast with large warehouse sizes and unoptimized queries, and real-world bills for active data teams routinely run $5,000-$20,000/month. The alternative is BigQuery's on-demand model — $6.25/TB queried with no warehouse-hour cost, which is cheaper for intermittent query patterns. For continuous, high-concurrency workloads, Snowflake's compute sizing is more predictable.

Product roadmap

Snowflake is building out its AI capabilities through Cortex AI (LLM functions in SQL, document AI, ML features), expanding Snowpark for ML workloads, and positioning as the AI Data Cloud — not just a warehouse but an execution environment for AI pipelines.