The open source Postgres platform scaling for the AI era

Paul Copplestone encountered a scaling wall in 2014 while using Firebase for a startup chat application. The Firestore database imposed a limit of one write per document every second. This constraint prevented his team from scaling chat messages. He built a real-time engine on top of Postgres using Elixir and Phoenix. He open-sourced this engine. In April 2020, he and co-founder Ant Wilson launched Supabase with a $100,000 pre-seed. They met through the Entrepreneur First program in Singapore and lived together in an Orchard penthouse for a year. They initially used the tagline "Real-time Postgres." This description told users nothing about the product’s purpose. They changed the headline to "The open-source Firebase alternative." He hit save. Within 72 hours, hosted databases jumped from eight to eight hundred.

The company now employs 350 employees across 40 countries.

The team builds tools for a new generation of developers. Copplestone’s experience with a marketplace in Southeast Asia and an office-management startup called Nimbus informed his approach to product-market fit. He focused on building a distributed team with an async workflow. This approach allows him to hire senior engineers who own their own roadmaps. They hit scale.

Comparing backend costs and architecture

Full-stack teams face massive cost differences when selecting a backend. At 100,000 monthly active users, the identity bill swings by more than $1,000 per month depending on whether the team chose Supabase, Firebase, or AWS Amplify.

Supabase is a managed PostgreSQL instance with a REST and realtime layer. It uses Postgres to enforce security through Row Level Security. This method places authorization logic inside the database. Firebase evaluates authorization through security rules. These rules exist in a separate JSON-like file. This separation can cause developers to get security rules wrong. AWS Amplify uses Amazon Cognito for identity. Cognito supports SAML and OIDC federation. You should evaluate your team’s familiarity with SQL before picking Supabase.

Firebase is a choice for mobile-first apps. Firestore is a document database organized around collections and documents. Google Cloud Storage powers Firebase Cloud Storage, which charges $0.026 per GB stored per month. Supabase uses S3-compatible object storage. Its costs include a 250 GB egress allowance in the Pro plan. AWS Amplify handles realtime through AppSync GraphQL subscriptions. Users pay $2.00 per million real-time update connection-minutes.

Supabase is an advantage for AI workloads. Adding pgvector allows users to join a similarity search against user permissions in a single SQL query. This prevents the need for a separate vector database. As of July 2026, Supabase Realtime Broadcast supports binary payloads. This upgrade helps developers building multiplayer applications.

AI agents and the rise of Multigres

AI coding agents now drive the majority of database deployments. Anthropic’s Claude Code is the largest contributor to new database deployments in 2026. This demand produced a 600% year-over-year increase in database creation. In June 2026, Supabase raised $500 million in a Series F round led by the Singapore sovereign wealth fund GIC, which brings their total capital raised to more than $1 billion. The company now holds a $10.5 billion valuation. This valuation doubled in eight months. Supabase serves 9 million developers and 250,000 customers.

The company saw 370% customer growth for its "Supabase for Platforms" tool over the last six months. This growth relies on "vibe coding" tools like Claude Code and OpenAI Codex. These tools allow developers to provision backends programmatically. The funding round included investors like Accel, Y Combinator, Craft, Felicis, Peak XV, Coatue, and Stripe. Salesforce Ventures joined as a new strategic backer.

The scale of AI workloads requires new infrastructure. Traditional single-instance Postgres deployments hit limits when AI agents spin up thousands of databases daily. Supabase released Multigres in preview to solve this. Multigres provides sharding and zero-downtime migrations under the Apache 2.0 license.

Will Multigres handle the next wave of autonomous agents?