Analytical SaaS is structurally over as AI models replace the query and analytics layer, destroying the profit pool.
AI coding agents are destroying the engineer-scarcity moat that justifies Snowflake's premium valuation
AI compresses Snowflake migrations 12mo to 1-2 and lifts win rates 75%, accelerating ARR and re-rating shares.
Infrastructure software undervalued: enterprises will store 10x the data in 3 years to defend and run AI.
Snowflake is in the AI token flow; more tokens means more database queries and a new AI multiple
Enterprises rip out Snowflake to fund GPUs, migrating to cheap Postgres-on-cloud; consumption growth stalls.
The counter-leg to Gerstner's app-layer short: data-platform companies reaccelerate because every agentic tool depends on the data and transformation that lives in their platforms, not in a thin layer on a CRUD database. Snowflake is the cleanest public name he stakes.
Agents turning Office artifacts into queryable backends grows the database layer faster than raw compute.