Supabase Buys Turso Because Agents Now Spin Up a Million Databases a Week
The number that justifies this deal: Supabase is launching over one million databases per week. Most of them aren't being created by people. Agents spin them up for prototypes, dashboards, explorations and throwaway apps. On Friday, Supabase announced it is acquiring Turso, the team that rewrote SQLite in Rust, to handle that flood.
The argument is about unit economics. Postgres is what you want once an app is real, but provisioning a dedicated Postgres machine for every agent experiment is absurd. Supabase's framing: "Agents should be able to create a database as easily as creating a file, with just as little concern about cost." Turso already built that. A single Turso server can manage millions of SQLite databases, loading each one when it's needed and suspending it when it isn't. Superhuman, Sauna.ai, CTO.new and Mastra already use it to give every agent its own database on demand.
So the combined pitch is a ladder. An agent starts with a near-free SQLite database. If the thing it built turns out to matter, there's a path up to Supabase Postgres without changing developer experience. Nothing changes for existing users of either product. Turso cofounders Glauber Costa and Pekka Enberg join Supabase, and Glauber Costa will lead what Supabase calls its "agentic infrastructure effort."
The bigger signal: infrastructure companies are now sizing their roadmaps around agent demand rather than human demand. Supabase says outright that database demand from agents will outpace the world's current capacity. Neon got bought by Databricks on a similar thesis last year. The database for agents is turning into its own category, and its defining property is cheap, disposable and many, not big and durable.
Link: supabase.com/blog/supabase-is-acquiring-turso
← Back to all articles
The argument is about unit economics. Postgres is what you want once an app is real, but provisioning a dedicated Postgres machine for every agent experiment is absurd. Supabase's framing: "Agents should be able to create a database as easily as creating a file, with just as little concern about cost." Turso already built that. A single Turso server can manage millions of SQLite databases, loading each one when it's needed and suspending it when it isn't. Superhuman, Sauna.ai, CTO.new and Mastra already use it to give every agent its own database on demand.
So the combined pitch is a ladder. An agent starts with a near-free SQLite database. If the thing it built turns out to matter, there's a path up to Supabase Postgres without changing developer experience. Nothing changes for existing users of either product. Turso cofounders Glauber Costa and Pekka Enberg join Supabase, and Glauber Costa will lead what Supabase calls its "agentic infrastructure effort."
The bigger signal: infrastructure companies are now sizing their roadmaps around agent demand rather than human demand. Supabase says outright that database demand from agents will outpace the world's current capacity. Neon got bought by Databricks on a similar thesis last year. The database for agents is turning into its own category, and its defining property is cheap, disposable and many, not big and durable.
Link: supabase.com/blog/supabase-is-acquiring-turso
Comments