In April 2020, Supabase hosted 8 databases. In June 2026, investors valued it at $10.5 billion.

Supabase's track record: from a broken chat feature to the default backend for AI agents.

Supabase's track record: from a broken chat feature to the default backend for AI agents.

Supabase is the open-source backend behind millions of apps: database, logins, file storage and APIs in one place.

It matters because it beat Google's Firebase without a big ad budget or enterprise deals, and today AI agents pick it by default.

It worked because of deliberate decisions about positioning, trust and shipping rhythm, and its biggest test is happening right now.

Below is the full story, act by act, plus a 90-day plan to apply the same levers in your startup.

Inside you'll find:

  1. The Origin: A Chat Feature That Kept Breaking (2015–2019)

  2. The Turning Point: One Tagline, 8 to 800 Databases in 3 Days (2020)

  3. The Build: Launch Weeks, Day Zero and a Three-Year Beta (2021–2024)

  4. The Explosion: When AI Agents Became the Customer (2024–2026)

  5. The Conflict: 16,000 Exposed Databases and a $1B Rival

  6. Steal the Playbook: The 90-Day Supabase Plan for Founders

1. The Origin: A Chat Feature That Kept Breaking (2015–2019)

Supabase didn't start as a database company. It started as a bug in someone else's startup.

Paul Copplestone grew up on a farm in New Zealand, left school at 18 and in 2015 co-founded ServisHero, a home-services marketplace in Kuala Lumpur. Later, through the Entrepreneur First program in Singapore, he met Ant Wilson, an engineer with three venture-backed startups behind him (Stacksync).

The spark came from Paul's next product, a workplace tool with chat. The team used Firebase for messages until they hit Firestore's limit of "one write per document per second." Postgres, the database they trusted, couldn't push updates to screens in real time.

So Paul built the missing piece himself, a real-time engine on top of Postgres, and open-sourced it. It got 62 upvotes on Hacker News. Not a launch, but a signal.

→ The best infrastructure ideas often come from a workaround a founder was forced to build. If you keep rebuilding the same plumbing, someone else probably is too.

2. The Turning Point: One Tagline, 8 to 800 Databases in 3 Days (2020)

In January 2020, Paul and Ant started Supabase with a $100,000 pre-seed and an empty repository. By April they had a working product and exactly 8 hosted databases. The site said "real-time Postgres." It described the product perfectly, and nobody cared.

Then Paul changed one line: "The open-source Firebase alternative."

A user posted it to Hacker News. It drew 1,120 upvotes, became the second most popular Launch HN at the time, and hosted databases jumped from 8 to 800 in three days (Basedash).

Hosted databases went from 8 to 1 million in four years.

Hosted databases went from 8 to 1 million in four years.

Two decisions turned that spike into a company:

  • They borrowed a known category. Every developer knew Firebase. "Open-source" told them the one thing that was different.

  • They chose a dense first market. They joined Y Combinator's Summer 2020 batch partly because the batch itself, 100+ venture-backed startups, was their ideal customer list.

"Product-market fit often is just like a product-positioning fit." Paul Copplestone, First Round Review

→ Before you rebuild the product, test a new one-line positioning against a category your buyer already understands.

3. The Build: Launch Weeks, Day Zero and a Three-Year Beta (2021–2024)

After YC, most startups lose their pace. Ant Wilson's fix: "pretend that we're starting the batch again" and recreate an accelerator inside the company, complete with its own Demo Day (Supabase blog). In the retro of the first cycle, someone asked, "why just have one launch?" In the last week of March 2021, the first Launch Week shipped 7 major features in 5 days, including paid pricing. Until then, Supabase had been completely free.

The rules were simple: fixed timeline, flexible scope, and the engineer who built each feature writes its launch post. The company credits this rhythm with 47% month-over-month growth in managed databases for 18 months. Copplestone says each launch still brings a roughly 20% bump.

Under the launches sat a deeper strategy: win day zero, then make leaving easy. Databases are rarely migrated, so the winner is whoever owns the moment a project starts. "The key metric is time to value," Copplestone says.

Where Firebase locked you in, Supabase did the opposite: standard Postgres, so your data is portable, and open source, so you can self-host. It also underpromised on purpose, keeping the beta label from December 2020 until April 2024 while developers shipped production apps on it (DevClass).

→ A launch is a distribution event, not a release. And for developers, a visible exit door is what earns the first commit.

4. The Explosion: When AI Agents Became the Customer (2024–2026)

Copplestone describes the AI wave in three phases. First, vector search inside Postgres. Then, at the end of 2024, Bolt and Lovable launched and both made Supabase their default backend. Then, from early 2026, coding agents like Claude Code and Codex started creating databases on their own.

Accel partner Arun Mathew put it bluntly: "We haven't seen a company grow at this pace, certainly in the database layer, ever before" (SiliconANGLE). The numbers:

  • Revenue run-rate went from about $30M at the end of 2024 to an estimated $170M by May 2026 (Sacra). Supabase doesn't publish revenue, so treat these as outside estimates.

  • Nearly 10 million developers, more than double the count of eight months earlier (TechCrunch).

  • More than 60% of new databases are now launched by AI tools, with Claude Code the largest single source.

Estimated revenue run-rate grew about 5.7x in 17 months.

Estimated revenue run-rate grew about 5.7x in 17 months.

Capital raced to keep up, ending with a $500M Series F led by GIC at $10.5B in June 2026 (Orrick).

Valuation multiplied 5.25x in 14 months.

Valuation multiplied 5.25x in 14 months.

Just as telling is what they refused to do. Supabase regularly turns down large enterprise contracts that come with custom demands; Copplestone admits it "feels very painful" (TechCrunch). And sales reps are paid "on the incremental uplift only." To estimate it, Supabase compares accounts that reply to outreach with those that don't: about 125% quarter-over-quarter growth versus 25%.

Accounts that replied to outreach grew about 5x faster than those that didn't (not a randomized comparison).

Accounts that replied to outreach grew about 5x faster than those that didn't (not a randomized comparison).

One caveat: customers who reply to sales are often already the most engaged, so part of that gap would have happened anyway. The idea is right; the measurement can be sharper, and the playbook below shows how.

→ In a product-led company, accounts grow on their own. If you pay sales on total revenue, you pay for growth that was coming anyway.

5. The Conflict: 16,000 Exposed Databases and a $1B Rival

Every advantage of the AI wave has a cost, and Supabase is paying it now.

The first warning came in March 2025. Replit engineer Matt Palmer scanned 1,645 apps showcased by Lovable and found 170 exposing user data, including names, emails, financial details and API keys. The cause was Supabase databases with missing or wrong access rules (Semafor).

In September 2026 it became systemic. Security firm UpGuard checked around 300,000 domains showing signs of Supabase use and found 16,326 databases with readable tables. Exposed data included names, addresses, phone numbers and, in some cases, passwords and authentication tokens (TechCrunch).

The security cost of AI-built apps, from one showcase to the whole ecosystem.

The security cost of AI-built apps, from one showcase to the whole ecosystem.

Supabase's position: projects are "secure by default," and security is a shared responsibility. But when an agent, not a developer, configures the database, nobody is reading the warnings. The company has added row-level security on by default for dashboard-created tables, weekly security advisor emails and automatic revocation of leaked keys (Supabase).

The second front is competition. In May 2025, Databricks agreed to buy Neon, a rival serverless Postgres, for about $1 billion, saying 80% of Neon's databases were already created by AI agents (TechCrunch). Supabase, for its part, launched Multigres alongside the Series F, a tool that Copplestone says will help companies "scale up to the size of OpenAI."

→ The lesson: the trait that made Supabase grow, being the easiest start, is now its biggest risk. Easy for agents must also mean safe by default, or the next "open-source alternative" will be pitched against you. (If you build on Supabase yourself, we covered how to configure it on purpose in Lovable gets you moving fast. It can also expose your users' data.)

What's next: the full 90-day plan to run Supabase's seven growth levers in your own startup, with step-by-step instructions, real examples from Cal.com, PostHog and Resend, tools, four copy-paste prompts and the four metrics that turn this playbook into your next fundraising story.

logo

Subscribe to CLUB to read the rest.

Become a paying subscriber of CLUB to get access to this post and other subscriber-only content.

Upgrade

Already a subscriber? Sign in

A subscription gets you:

  • Everything included in Premium.
  • Early access to new tools and the specialized AI and product-market fit newsletter.
  • One-on-one feedback on your business, with actionable guidance on strategy, positioning, and growth
  • Masterclasses with Pablo Fernandez Alvarez

Reply

Avatar

or to participate