Paul Copplestone and Ant Wilson founded Supabase in 2020 with a deliberately boring pitch: a developer-friendly database, sold into a market Amazon and Google already owned.
Six years later, most of the customers placing orders are not people.
More than 60% of the new databases spun up on Supabase are now started by an AI coding tool rather than a human, Copplestone said this week. Database launches grew 600% in a single year. On Thursday, investors put a number on that shift: half a billion dollars in fresh funding at a $10.5 billion valuation, roughly double what the company was worth in October.
The round is one of the clearest financial bets yet on vibe coding, the practice of building software by describing it to an AI in plain language and letting the model write the code. Every app those models generate needs somewhere to store data, sign in users, and run in production. Supabase has quietly become the place a lot of that machine-written software lands.
Claude Code Became Supabase's Biggest Customer
Supabase sells the unglamorous layer underneath an application: a managed Postgres database, plus user authentication, file storage, APIs, real-time updates, and vector search for AI features. The pitch, printed on the company's own site, is "build in a weekend, scale to millions."
For years the people doing the building were human developers. That has changed fast. AI coding assistants such as OpenAI's Codex and Anthropic's Claude Code can now scaffold a working app from a text prompt, and when they need a backend, they increasingly reach for Supabase on their own.
Copplestone put hard numbers on it in announcing the round.
"Demand for Supabase is exploding. Our user base has more than doubled since the Series E, and we've seen a 600% increase in databases year-over-year. Claude Code is the largest contributor since the start of the year. Agents are now deploying the majority of databases on our platform." — Paul Copplestone, CEO and co-founder, Supabase (Series F announcement, June 4, 2026)
Nearly 10 million developers now build on Supabase, the company says, more than double the figure from eight months earlier. It counts 250,000 customers and roughly 350 employees. Supabase for Platforms, the product that hosts other companies' AI app builders, grew its own customer base 370% in six months.
That a single coding agent is the biggest source of new databases is the whole thesis in miniature: the customer is now often a model, and the model works around the clock.
A Postgres Backend Built for the Vibe Coding Era
From the start, Supabase was built as an open-source alternative to Google's Firebase, sitting on top of PostgreSQL, the decades-old open-source database that engineers already trust. The bet was that developers wanted Firebase's convenience without being locked into a proprietary stack.
That open-Postgres foundation turned out to suit the AI moment. A model generating an app does not want to learn a bespoke system; it wants standard Postgres, standard SQL, and a single platform that handles the rest. Supabase fits that shape.
Singapore's sovereign wealth fund GIC led the Series F, which valued the company at $10.5 billion. Total capital raised now passes the one billion dollar mark, and the valuation has roughly doubled at each of its last two rounds.
| Date | Valuation | Lead investor(s) |
|---|---|---|
| April 2025 | $2 billion | Accel |
| October 2025 | $5 billion | Accel, Peak XV |
| June 2026 (Series F) | $10.5 billion | GIC |
The company competes with database offerings from MongoDB and the cloud services sold by Amazon.
The Number Investors Are Buying Is Usage, Not Revenue
Here is the catch in the valuation: Supabase has not disclosed its revenue. It declined to give a figure at the October round, Fortune reported, and named none this week either. The doubling ran on usage, the 600% jump in launches and the developer count, not on a profit-and-loss statement.
Investors are betting that the apps agents spin up tonight become the paying workloads of next year. It is the same usage-first logic now pricing AI coding companies like Cursor, whose valuation doubled on revenue growth rather than profit. Accel partner Arun Mathew, whose firm has backed Supabase through multiple rounds, called the pace unmatched.
"We haven't seen a company grow at this pace, certainly in the database layer, ever, ever before." — Arun Mathew, Partner, Accel (CNBC, June 4, 2026)
Copplestone said the new money has three jobs: supporting growth, accelerating Supabase's open-source Postgres tooling, and providing liquidity for employees. That last item typically funds a secondary share sale, where existing shares change hands rather than new cash reaching operations.
The Catch: Agent-Built Databases Tend to Get Thrown Away
A database "launch" counts a machine starting one up. It does not count whether anyone keeps it. And the closest precedent suggests a lot of agent-created databases are disposable.
When Databricks bought the serverless-Postgres startup Neon in May 2025 for roughly $1 billion, Neon disclosed that the share of its databases created automatically by AI agents, rather than by humans, had climbed from 30% to more than 80% as the product matured. Those instances spun up in under half a second and were designed to be branched, forked, and discarded. They were cheap precisely because most of them did almost nothing before being thrown away.
Supabase's headline metric counts that same kind of activity. A 600% rise in launches is not a 600% rise in production applications, and the company has not said how many of its AI-started databases turn into paid, persistent workloads.
The durability of vibe-coded software itself is contested. Andrej Karpathy, who coined the term "vibe coding," framed it from the start as suited to "throwaway weekend projects." Developer Simon Willison, who has tracked the practice closely, has called vibe coding your way to a production codebase "clearly risky," because real engineering means evolving systems you actually understand. Red Hat's developer group wrote in February that "so many vibe-coded projects hit a wall around the three-month mark," once a codebase outgrows anyone's ability to hold it in their head.
There is a concentration risk, too. With Claude Code as the single largest source of new databases, Supabase's growth curve is now tied to the trajectory of one outside vendor's coding agent, and to the staying power of a programming style barely two years old. Anthropic, which makes Claude Code, confidentially filed to go public last week.
Multigres Is a Bet on the Opposite Problem
Alongside the funding, Supabase shipped an early version of Multigres, an open-source scaling layer for Postgres, as a v0.1 alpha the company calls "ready to try, but not yet production-ready." It is built by Sugu Sougoumarane, co-creator of Vitess, the system that scaled YouTube's databases. The goal, Copplestone said, is to let companies scale "up to the size of OpenAI or even larger" without migrating off Postgres.
That points Supabase at the exact opposite of the workload it is growing on. Multigres aims at the largest, most permanent Postgres deployments in the world. The 600% launch figure runs on the smallest and most ephemeral: databases that AI tools open in bulk and may abandon by morning. Which of those two underwrites a $10.5 billion price tag is the question the next round, or the first disclosed revenue figure, will answer.
The Bottom Line
Supabase is one of the first companies to be repriced primarily on what AI agents do, not on what humans buy. The case for the valuation is real: every piece of machine-written software needs a backend, Supabase has become a default, and adoption is compounding faster than anything its investors say they have seen in the database business. Nearly 10 million developers and a 600% jump in launches are not nothing.
The case against it is just as concrete. The company is priced on usage it will not yet convert into a public revenue number, its single biggest source of growth is another company's coding agent, and the one detailed precedent for agent-built databases says most of them are throwaways.
Copplestone is now selling two futures at once: a flood of disposable databases that machines open by the thousand, and a heavyweight scaling engine for the few apps that survive to become the next OpenAI. The bet behind the $10.5 billion valuation is that enough of the first quietly turns into the second. Nobody has disclosed the conversion rate yet, which is the only number that will settle whether this was vision or a vibe.
Sources
- Database startup Supabase raises $500 million at a 10.5 billion valuation (CNBC, June 4, 2026)
- Supabase Doubles to $10.5 Billion as Agents Deploy Most of Its Databases (Implicator.ai, June 4, 2026)
- Supabase hits $10.5B valuation after a 500M funding round as vibe coding fuels explosive growth (Tech Startups, June 5, 2026)
- Supabase Raises $500M at 10.5B to Accelerate Lead in Agentic Infrastructure (PR Newswire / Supabase, June 4, 2026)
- Databricks agrees to acquire Neon to help developers deliver AI systems (Databricks, May 2025)