Supabase to Acquire Turso, Raises $150 Million to Build the Database Layer for AI Agents

At 09:00 ET on 2 October 2026, Supabase published a single release carrying two pieces of news. The first: it is acquiring Turso (opens in a new tab), a database infrastructure company. The second: it has raised $150 million in new primary funding, led by Singapore’s sovereign wealth fund GIC, with Alphabet’s growth-stage fund CapitalG, along with IronArc and SquarePeg, also participating.

The two events are related in timing and strategic logic, not in structure. The $150 million is equity capital into Supabase itself. The Turso acquisition is a separate purchase of a company, and Supabase has not disclosed what it paid. That silence is unremarkable for a private-to-private technology acquisition of this size; it is the funding round, not the acquisition price, that carries disclosed numbers here.

The headline numbers

• –Funding. $150 million in new primary equity, arriving roughly four months after Supabase’s $500 million Series F.
• –Lead investor. GIC, with CapitalG, IronArc and SquarePeg also participating in the round.
• –Acquisition. All of Turso, the company and its team, with terms undisclosed.
• –Use of proceeds. The release states the round will also fund employee liquidity, something the founders describe as important to being, in their words, the best company for engineers.

Paul Copplestone, Supabase’s co-founder and CEO, framed the round as recognition of a growth rate rather than a rescue or a pivot: "The past year has been the fastest growth in our history by a wide margin." He credited the investor group, and specifically CapitalG’s ties into the broader Alphabet ecosystem, as valuable for what comes next.

Derek Zanutto, general partner at CapitalG, offered the investor’s framing of the same moment: "Supabase has established itself as the critical open infrastructure layer powering the AI revolution. We’re proud to back one of the fastest growing private database companies on the planet as Paul and the team empower millions of developers and agents to build at unprecedented speed."

Both parties are describing the same underlying fact from different sides: Supabase’s own usage data shows growth concentrated in a kind of demand, agent-created databases, that its existing Postgres-based platform (opens in a new tab) was not originally built to serve at this volume. That is the thread that connects the funding round to the acquisition, and it is the entire rationale for buying Turso.

A traditional database platform, Supabase’s own included, is built around the assumption that a database is provisioned for a project, a company or an application, and that it will be used by some number of human end users and backend services over a long lifetime. Supabase built an entire company on making that provisioning step fast and the resulting database fully featured, with auth, storage, edge functions, realtime and vector search (opens in a new tab) included by default.

Agentic workloads break that assumption in a specific way. An AI coding agent or an autonomous service might need its own database for a single task, a single customer, or a single user session, and it might need thousands or millions of those databases created and torn down continuously rather than once. Supabase’s release puts that shift in its own numbers: more than 4 million new databases created every month platform-wide, with roughly 70% of those now created by agents or AI-driven tools rather than a person clicking through a dashboard, up from a point in June 2026 when the company had already reported 600% year-over-year growth in total databases created.

Why you cannot just provision a Postgres instance per agent

A full Postgres instance, even a lightweight one, still carries real per-database overhead: memory reserved for the server process, connection pool slots, background workers and monitoring. Multiply that overhead by millions of concurrently live databases and the economics invert: the infrastructure cost of holding all those idle-but-ready databases open can exceed the value of the work any single one of them does.

Turso’s starting point was different. It rebuilt SQLite, a database engine designed from the outset to be embedded and lightweight, adding the capabilities SQLite traditionally lacked for server use: concurrent writers through multi-version concurrency control, asynchronous I/O, and built-in sync back to a central cloud layer. The result, according to Turso’s own description, is an engine "lightweight enough to multiply, fast enough to run anywhere," that stays a drop-in-compatible superset of SQLite’s SQL, schema and query behaviour.

That lets Turso Cloud, the hosted product built on top of the engine, offer something a conventional managed-Postgres platform struggles to match at scale: instantiating a new, isolated database in the time it takes to make an API call, with a cost structure that holds up even when most of those databases are small, short-lived, or barely used. Turso also ships features that matter specifically for agent and multi-tenant workloads out of the box, including full-text search with BM25 ranking, native vector embedding storage and similarity search, per-database encryption at rest, and change data capture for audit logging of every mutation.

Public cloud and BYOC, not either-or

The release is specific that Turso’s architecture "enables the provisioning of on-demand per-agent databases both in the public Cloud and BYOC," bring-your-own-cloud, where the infrastructure runs inside a customer’s own cloud account rather than Turso’s. That detail matters for the kind of customer Turso names in the release: Superhuman, Sauna.ai, CTO.new and Mastra, companies building AI products where regulated or security-conscious end customers may require that per-agent data never leaves their own network boundary.

Four months is an unusually short gap between a $500 million Series F and a further $150 million raise, even in the current AI infrastructure financing environment. Supabase has not characterised this round as an extension, a bridge, or a down round of any kind; the release frames it purely as new capital arriving because growth accelerated further after the Series F closed, not because of any shortfall.

Who is in the round, and why it reads as a signal

• –GIC. Leads the round. GIC is one of the largest sovereign wealth funds globally, and a GIC-led growth round is typically read as a vote on durability and scale, not just near-term momentum.
• –CapitalG. Alphabet’s independent growth fund. Copplestone specifically called out CapitalG’s "connections to the broader Alphabet ecosystem" as valuable going forward, which reads as more than a passive capital commitment.
• –IronArc and SquarePeg. Round out the syndicate. The release does not break out amounts by investor.

The release also states plainly that the round "will provide liquidity for employees," which the founders describe as "an important and vital part of being the best company for engineers." That is a deliberate, disclosed use of proceeds: at this stage of a company’s life, a secondary liquidity component in a primary round is usually there to let early employees and some early investors realise partial gains without forcing a sale of the company, which keeps talent retention intact through a long private tenure.

Two integration decisions stand out in the release, and both point toward continuity rather than disruption. First, the Turso platform is not being wound down or merged overnight: "The Turso platform will continue operations, with a clear graduation path into the rest of the Supabase ecosystem." Second, Turso’s leadership is not exiting; it is moving into a defined role inside Supabase.

Leadership after close

• –Glauber Costa. Turso’s founder joins Supabase as Head of Agentic Services, a title that puts him in charge of exactly the capability Supabase bought his company for.
• –Pekka Enberg. Turso’s co-founder joins alongside Costa, along with "the rest of the Turso team," per the release.

That quote is worth reading closely, because it reframes the deal from Turso’s side as scale-seeking rather than exit-seeking. Costa is not describing a team cashing out; he is describing a company that hit a ceiling on how many databases it could plausibly reach on its own distribution, and bought itself a route past that ceiling by joining the platform with 13 million developers already building on it.

The release frames the cultural fit in similar terms on both sides: "Both companies have a strong focus on open source, developer experience, a willingness to tackle hard problems, and a shared understanding for how infrastructure will evolve to support the future of agentic AI." That is standard acquisition language, but it is also consistent with the public track record of both teams: Supabase has built its entire platform in the open, and Turso’s rewritten SQLite engine and client libraries are likewise open source projects with existing external contributors and dependents.

A "graduation path" rather than a hard migration deadline is a meaningful integration choice. It avoids forcing Superhuman, Sauna.ai, CTO.new, Mastra and Turso’s other customers to re-platform under time pressure immediately after an acquisition, a pattern that has damaged customer trust in other infrastructure deals. It also buys Supabase time to decide exactly how deep the integration between a Postgres-based platform and a SQLite-based one should go, rather than committing to an architecture before the ink is dry.

Supabase is not the only company racing to make "spin up a database instantly, cheaply, and in enormous volume" a core product capability. Every modern backend-as-a-service and database platform is watching the same shift: coding agents and autonomous software are becoming database customers in their own right, not just a new kind of user hitting an existing database through an API. Owning Turso’s architecture, rather than licensing it or building an equivalent in-house, gives Supabase a two-sided advantage: it removes a potential competitor from the market and compresses years of specialised engineering into a single acquisition.

What Supabase is actually buying

It is worth being precise about what transfers in a deal like this, because the headline, "Supabase acquires Turso," understates how much of the value sits in people and operating experience rather than code alone. Supabase gets: a production-proven engine for high-volume, low-cost database provisioning; a team led by Costa and Enberg who have already solved the hardest parts of running that engine at scale, including for BYOC deployments; and a small set of reference customers, Superhuman, Sauna.ai, CTO.new and Mastra, already running real agentic workloads on exactly the architecture Supabase now needs to scale.

The defining value driver is architectural: Supabase’s core identity and most of its revenue sit on Postgres, and Turso’s engine is a SQLite-compatible system. Running both well, long-term, with a clear story for customers about which engine to pick, is a genuinely interesting product problem, and Costa’s new title, Head of Agentic Services rather than, say, Head of Turso, suggests Supabase intends to solve it by organising around the workload, not the database engine.

Because the companies disclosed a strategic rationale and a funding round but not a purchase price or an integration timetable, most of what happens next will be visible in product behaviour rather than further press releases. Three things are worth watching over the coming two to three quarters.

• –Whether Turso pricing changes for new customers. A straightforward signal of whether Supabase intends to grow Turso’s existing customer base independently or primarily route new agentic demand through Supabase’s own front door.
• –How database-per-agent is packaged inside Supabase. Whether it ships as a distinct Supabase product, a mode inside existing Supabase projects, or stays a separate Turso-branded offering during the "graduation" period.
• –Whether other platforms respond with their own acquisitions. Database-per-agent infrastructure is now a proven-enough category that other developer platforms with large existing user bases have a motive to either partner with a remaining independent provider or make a similar purchase of their own.

For now, the clearest fact in the entire announcement is the one both sides chose to make public: not the price, but the scale. Supabase is adding four million databases a month and expects the agent share of that number to keep climbing; Turso says its founding goal has always been to serve upward of a billion databases. Put the two together and the acquisition reads less like a typical tuck-in and more like two companies independently arriving at the same forecast for how agentic software gets built, and deciding to meet it together rather than compete over who gets there first.