Warehouses Absorb Analytics to Monetize Compute
George Xing, co-founder and CEO of Supergrain, on the future of business intelligence
This reveals why the warehouse vendors are naturally pulled up the stack. If Snowflake or Databricks can get customers to run more dashboards, metric layers, ML jobs, streaming pipelines, or embedded data apps on their platform, they increase the amount of paid processing that runs through their system. In practice, that means every new workload is not just a feature win, it is more metered compute revenue and deeper product lock in.
-
Snowflake charges for compute through credits consumed by virtual warehouses and serverless features, with usage rising as warehouses get larger or run longer. That makes products that trigger more queries, refreshes, and application traffic directly additive to revenue.
-
Databricks uses the same basic model with DBUs, a metered unit of processing power. Its push from data engineering into SQL, BI, AI, and apps follows the same logic, because one customer can start with one workload and then expand spend as more teams run jobs on the platform.
-
That is why the semantic layer and metrics layer are strategically contested. Standalone tools like Preql and dbt create value by defining business logic once across tools and even across multiple warehouses, while warehouse vendors want those workflows native so more usage stays inside their own walls.
The direction of travel is clear, more of the analytics stack will be absorbed into the warehouse and lakehouse platforms, especially the pieces that cause repeated queries and ongoing model execution. Independent tools will keep winning where cross cloud support, neutral governance, or best of breed workflow matters more than keeping all compute on one platform.