Supabase acquires Turso to build on-demand database infrastructure for AI agents
- Supabase announced on 2 Oct 2026 that it is acquiring Turso, the company behind the Rust rewrite of SQLite, to build database infrastructure sized for AI agents that spin up databases on their own.
- Supabase already launches more than 1 million databases per week and argues that agent-driven demand will outpace the world's current capacity to support it.
- Turso's architecture lets a single server manage millions of databases, loading them on demand and suspending idle ones; Superhuman, Sauna.ai, CTO.new, and Mastra run on it, either in Turso Cloud or their own clouds.
- Turso keeps operating and keeps working on SQLite while Supabase stays on Postgres; the post says existing users see no change and that Turso gets a path into the Supabase ecosystem as workloads grow.
- Glauber Costa and Pekka Enberg join Supabase with the rest of the Turso team, and Costa will lead the agentic infrastructure effort.
Hacker News opinions
Always been a fan of Supabase but I did not expect this one. Curious to see what comes out of it.
I use Turso for a small project and it works well. Just hope this does not change anything for me.
I was always a bit worried that Turso's future was tied to the company's success, and it actually changed my choices. I picked plain sqlite over Turso a few times because of that. Now I'll probably go with Turso.
Wait, are you seriously preferring a vibe coded project over the war tested sqlite?
As a Turso customer, ugh, I do not want this at all.
Write me a message directly or hit me up on Discord. We are very excited to be part of Supabase.
We use Turso for async stuff in Rust projects and the performance has been very good. Looking forward to what comes out of this.
I hope they put more resources into fixing issues. There were multiple attempts to add it to ClickBench and each time new bugs turned up. It should not be several times slower than SQLite.
That ClickBench thread is wild. It was loading data for a week and the speed dropped to 4 KB per second, so it would take years. A year later it had loaded maybe 1 percent. Even with fsync removed it does not work. I know the codebase was heavily written by LLMs but the execution cannot be this bad. Why would Supabase buy a product that cannot even load the dataset?
I do not feel good about this. I really want SQLite, and the Turso rewrite of it, to win for the tiny apps people will build with LLMs, on desktop or serverless. The core tech is open source so that part is fine, but will it stay? I guess the hosting was not making enough revenue.
Turso will continue operating, with a clear path into the broader Supabase ecosystem as workloads grow.
Why would anyone pay for Turso when the core of the software is free and licensed permissively under MIT? You answered your own question.
I would guess it is more that the founders want to exit before AI replaces them, going by the doom and gloom narrative in tech right now.
Turso CEO here. We were doing fine. This is a strategic acquisition and I fell in love with the vision Supabase had for how they would use Turso going forward. More to come.
Open source never pays the bills, and users here never pay for open source either.
Our cloud pays the bill just fine though.
This actually makes a lot of sense. Rare win-win-win for Supabase, Turso, and the customers.
Why would you rewrite something widely considered one of the best written and tested pieces of software in the world?
That is how you push the field forward. Good is not good enough.