Supabase Raises $150M, Buys Turso for AI Agent Databases

Roughly seven out of every ten new databases created on Supabase right now aren't being set up by a person typing commands into a terminal. They're being spun up by AI agents — software that writes its own code, decides it needs somewhere to store data, and just creates a database to do it. That one statistic is why Supabase just raised $150 million and bought a smaller database startup called Turso.

What actually happened

Supabase, the open-source alternative to Google's Firebase that lets developers add a database, authentication, and file storage to an app without building a backend from scratch, announced the funding round earlier this month. It was led by GIC, Singapore's sovereign wealth fund, with Alphabet's growth-stage investment arm CapitalG, along with IronArc and SquarePeg, also putting money in. Notably, this comes just four months after Supabase closed a $500 million Series F — a later-stage funding round, usually larger than the one before it, that companies raise once they've already proven the business works and just need fuel to scale faster.

Alongside the raise, Supabase acquired Turso, a startup built around SQLite — a lightweight, file-based database engine that doesn't need a dedicated server running around the clock, which makes it fast to create and just as fast to throw away. Turso's founder, Glauber Costa, is joining Supabase as part of the deal, and Turso will keep running as its own product for now. The acquisition price wasn't disclosed.

Why a database company is suddenly about AI agents

None of this makes much sense unless you accept the premise driving it: a growing share of code today isn't written by developers at all, it's written by AI coding agents acting on a developer's behalf, and those agents increasingly need to create and tear down databases on their own, sometimes for a task that lasts a few minutes. A traditional database that takes time to provision and costs money to keep running doesn't fit that pattern. Something built to spin up instantly, do its job, and disappear does.

• Supabase's own numbers suggest the majority of new databases on its platform are now agent-created rather than human-created.
• Turso's SQLite-based architecture is built for exactly this — quick to start, cheap to abandon.
• Part of the new funding is also going toward employee liquidity, letting early staff cash out some equity — a sign the company is settling in for a longer runway rather than an imminent sale.

The database used to be the slow, careful part of building software. Agentic infrastructure is betting that it now has to be the disposable part.

Why this matters to Indian developers and startups

Supabase has a disproportionately large footprint among Indian developers — it's the go-to open-source Firebase substitute for bootstrapped founders, hackathon teams, and early-stage SaaS startups that want a generous free tier and don't want to manage their own servers. A lot of the "vibe-coded" apps and AI-agent-built prototypes coming out of Indian college hackathons and indie-hacker circles already run on it. As that trend scales up, it raises a genuinely India-specific question: under the Digital Personal Data Protection (DPDP) Act, companies are expected to know what personal data they hold and where it lives. If an AI agent is autonomously spinning up and deleting databases on a founder's behalf, keeping track of what data existed, for how long, and whether it ever needs to be reported or deleted becomes a lot harder — a compliance gap that most small Indian teams haven't had to think about yet.

For a sense of how this kind of infrastructure fits together, Code24's explainer on vector databases and how AI systems store information is a useful companion read.

The bigger picture

This deal is a small data point in a much larger shift: infrastructure companies across the board are re-architecting themselves around agents as the primary user, not developers typing commands. Supabase betting $150 million and an acquisition on that idea is a signal that the shift isn't theoretical anymore — it's already showing up in usage numbers. The next interesting question isn't whether agents will keep creating their own databases. It's who ends up responsible for cleaning up after them.