If your agents spin up their own databases on Turso, the price of that design now belongs to Supabase, and neither company has said what it will be. Supabase announced on October 2, 2026 that it is acquiring Turso, the SQLite company whose cloud lets you create thousands of small databases on demand. The announcements say Turso "will continue operating." They name no price, no close date, no SLA and no commitment on Turso's current plans. Until they do, the terms you have in writing today are the only ones you can count on.
The deal arrived with money attached. Supabase raised $150 million led by Singapore's GIC, with Alphabet's CapitalG, IronArc and SquarePeg joining, and the purchase price was not disclosed. Ventureburn reports that the round follows a $500 million Series F four months earlier, and that Supabase says 70% of new databases on its platform are now created by AI tools and coding assistants. That number is Supabase's own observation of its platform and has not been independently checked. It still explains why a Postgres company wants a SQLite one.
What Supabase Actually Bought
Supabase bought a team and a pattern for giving every agent its own database. Database-per-agent is an architecture in which each agent, task or tenant gets a dedicated, isolated database that is created on demand and costs almost nothing while idle. SQLite suits it because a database is a single file. Postgres, which Supabase is built on, is a server process with compute attached, so each new database carries a running cost.
Supabase's post is explicit about the split: Supabase keeps building around Postgres, Turso keeps working on SQLite, and together they plan to build "something new that will bring the Supabase experience to every agent." Turso founder Glauber Costa leads that effort. Turso's own post says he becomes Head of Agentic Services at Supabase, with cofounder Pekka Enberg and the rest of the team joining too.
Supabase CEO Paul Copplestone framed the demand this way, per SiliconANGLE: "Agents are spinning up millions of databases to power the prototypes, explorations, dashboards, and apps they're building." Supabase's post says it is already launching over one million databases per week.
What the Announcements Promise, and What They Leave Out
The only commitment in writing is that current workloads keep running. Turso's post says: "Turso keeps running. Your databases, APIs, and workflows continue as they do today." It adds that "Turso Database remains open source and actively developed," and tells customers to "expect deeper integration between Turso and Supabase in the coming months."
Read the gaps as carefully as the promises. Neither post mentions:
- Turso Cloud pricing, or whether the current plans stay on sale to new customers
- The free tier
- Uptime or support SLAs
- When the deal closes, or whether it already has
- How the promised move from SQLite to Postgres will work, what it costs, or what data it carries
The migration story is a single sentence in Turso's post: "When SQLite isn't enough, you can move into standard Postgres through the Supabase ecosystem, and all the way to petabyte scale with Multigres." No tool, timeline or price comes with it.
Why the Price List Is the Part Most Likely to Move
The two companies price databases in incompatible ways, and one of those models will have to give. Turso's pricing page, checked October 4, 2026, sells unlimited databases on its Developer plan at $4.99 a month (the annual-billing rate; monthly billing costs $1 more), with 9GB of storage and 2.5 billion monthly row reads included. The free plan allows 100 databases. Scaler is $24.92 a month and Pro $416.58 a month on the same annual basis; Pro adds SSO, BYOK encryption, HIPAA and SOC 2. Enterprise is custom and adds bring-your-own-cloud deployment.
Supabase, on its pricing page checked the same day, gives the free plan two active projects that pause after a week of inactivity. Pro is $25 a month with $10 of compute credit, enough for one Micro instance, and each additional project adds its own compute cost. Uptime SLAs appear only on Enterprise.
A billing model built on per-project compute cannot carry millions of agent-created databases at Turso's rates, and a $4.99 plan with unlimited databases is generous by any Postgres host's standard. That is our inference; neither company has said it. If it holds, the integration both posts promise will eventually need one price list, and you have no say in which one it resembles unless you get your current terms on paper.
Supabase's own billing has moved before. An independent analysis by Layerbase, citing Supabase's GitHub discussions, traces a shift to per-project compute charges in August 2023, a $4 monthly IPv4 add-on in January 2024 and hourly billing for add-ons in August 2024. Its argument is that the $25 headline price stayed fixed while the meters underneath it multiplied. That is one critic's reading, but the dated changes are checkable.
What Happened the Last Time Turso Changed Course
Turso has already shown how a "nothing changes" promise works in practice. In January 2025 it discontinued edge replicas for new users and removed multi-database schemas and ATTACH for new users, while telling paying customers their "existing production workloads will continue running exactly as they do today." Free-tier users were moved to AWS; paid users were encouraged but not required to follow.
The pattern is the one to plan for here. Existing workloads were protected, and the product sold to the next customer changed. If you plan to add Turso databases, regions or features next year, the current pledge may not cover them.
The engine underneath is also in motion. Turso's GitHub repository, MIT-licensed, says the Rust rewrite of SQLite "replaces libSQL as our intended direction," has "not yet reached 1.0," and advises users to "keep independent backups" until it does. If you self-host libSQL, you are already on the branch the company has said it is moving away from.
The closest precedent for the acquirer side is friendlier. Databricks agreed to buy Neon in May 2025, citing the same signal: over 80% of Neon databases were created by AI agents. Three months later Neon moved to usage-based pricing with no monthly minimum and a larger free tier. Acquirers do sometimes cut prices. You still want the floor in writing before anyone decides.
Who Should Act, and Who Can Wait
Teams with production agent workloads on Turso Cloud should act now; teams evaluating either platform can wait for the integration details. The difference is leverage. Today you are a customer Supabase wants to keep through a public transition. Six months after close, you are one line in a consolidated price list.
If you already run Supabase and never touched Turso, nothing changes for you yet. Watch for whether the promised "Supabase experience" for every agent arrives as a new SKU or as a change to how projects are billed. Self-hosters on Turso Database or libSQL face a governance question more than a pricing one: an open-source project owned by a company with a different primary engine needs a fork plan you have actually tested.
This Week:
- Export your Turso database inventory: count, size and creation rate per tenant or agent. That number decides how exposed you are to any change in how databases are counted.
- Take an independent backup of every production Turso database outside Turso Cloud, which the project's own README already recommends.
- Ask your Turso account contact, in writing, for the deal's expected close date and whether your current plan stays available to you after it.
This Month:
- Request a price lock on your current plan for at least 12 months after close, with 90 days' written notice of any change to how databases, rows or syncs are metered.
- If you are on Developer or Scaler, check which features you rely on (IP allow lists, point-in-time restore windows) and get them named in any amendment, since Turso has removed features for new users before.
- Run a test export of one representative database to plain SQLite and to Postgres, and time it. The "path to Postgres" will cost you something; find out how much before the vendor prices it.
Before Renewal:
- Add a change-of-control clause that lets you terminate without penalty if pricing, SLA or data location changes after the acquisition closes.
- If you are choosing between Supabase and Turso for a new agent backend, ask Supabase for the combined roadmap and price sheet in writing, and decline to sign a multi-year commitment until it exists.
The Bottom Line
Postgres hosting has gone through this cycle once already: Databricks bought Neon on the same agent-created-database argument, and Supabase is now making that bet with SQLite added. The technology case is sound, and the team that built Turso is staying together. What is missing is any written statement about what the combined product will cost or how long Turso's current plans will last, and Turso's own January 2025 changes show that a promise covering existing workloads leaves new ones open.
Get your database count, your backups and your price lock done before the close date is announced.
Continue Reading
- Vector Database Pricing: Only pgvector Publishes a Rate
- Baseten Buys Blaxel, Whose 'Keeps Running' Pledge Has No End Date
- Zep vs Mem0 vs Postgres: Buy the Memory Store Last
- Bending Spoons Bought Airtable. Reprice Before It Closes.
- AWS Didn't Buy DuckDB. It Hired Its Board Majority.
- Pinecone vs Weaviate vs pgvector: Stay on Postgres