Build2 publishers2 min readPublished
Supabase agrees to buy Turso to give each AI agent its own SQLite-compatible database
Supabase raised $150 million and agreed to buy Turso, whose SQLite-compatible engine gives AI agents large numbers of small, isolated databases. The deal rests on Supabase's own count of agent-created databases, and Turso's customers now build on that bet.
The Engineer · Build desk

What happened
- Turso founder Glauber Costa joins Supabase as Head of Agentic Services, along with co-founder Pekka Enberg and the rest of the Turso team.
- Supabase says Turso will keep operating, with a path into the broader Supabase platform.
- The announcement did not state the acquisition price or the timing and conditions for closing.
- The new round comes four months after a $500 million Series F that GIC also led, at a $10 billion pre-money valuation.
Compiled by The EngineerSomething wrong?How this is made
Why it matters
- capability Supabase can now offer a per-agent database next to the authentication, storage and vector search its Postgres platform already provides.
- decision A team that picks Turso for a new agent project now gets a product whose roadmap will be set inside Supabase.
- exposure Giving every agent its own database means more access settings to configure, and UpGuard's September 25 scan had already found 16,326 Supabase-linked databases exposing readable tables.
According to its developers, Turso treats each database as a lightweight file instead of a long-running server process [7]. Supabase says this lets Turso create databases on demand, in Supabase's public cloud or in a customer's own cloud, without a dedicated machine for each one [8]. The cloud service is diskless and keeps the write-ahead log on S3 [10]. The whole design is aimed at cost per database. Turso wrote: "Each agent deserves its own isolated database and the economics have to work at millions of databases." [11]
In my view the rewrite is the hardest engineering in the deal. Turso says SQLite's single-writer limitation held it back for server workloads [9]. The company rewrote the engine from scratch as open source, in Rust, and removed that limit so it supports concurrent writes [7][10].
The demand case comes from Supabase's own numbers. It says agents or AI-driven tools create 70% of new databases on its platform, out of 4 million added each month [4]. If both figures count the same population, agents create about 2.8 million databases a month there [1]. Supabase has not said how it counts users or databases [5]. It also has not said whether its method changed between the nearly 10 million developers it reported in June and the more than 13 million it reports now [6]. That gap is about 3 million developers in four months [2].
I think those figures support a narrow claim. On Supabase, most new databases are now created by software. Both announcements rely only on Supabase's numbers, so they cannot show that agents now set backend database design across the industry.
Turso lists Superhuman, Sauna.ai and Mastra among the teams building agents on it [12]. For those teams the owner changes first. The bigger change is the one Turso promises: "a seamless path to Postgres when an application outgrows them" [16]. RuntimeWire notes that developers still need details on how Turso's product, cloud service and open-source projects will fit alongside Supabase's Postgres-centered platform [19][18].
What to watch
- A closing date and terms for the Turso acquisition, which the announcement left out.
- Any change to Turso cloud pricing or bring-your-own-cloud terms once the product sits inside Supabase.
- The first product from Costa's agentic services group, and whether it runs Turso databases under Supabase authentication.