Serverless Postgres — the developer-first branchable database, now a Databricks company
HQ: San Francisco, CA (Databricks company; R&D across US/Netherlands/Germany/Israel) | Founded: 2021 | Employees: ~160-180 total, ~120 engineers | Stage: Acquired by Databricks (May 2025, ~$1B reported) | Website: https://neon.com
Neon is the serverless Postgres company founded in 2021 by veteran database engineers and incubated at Khosla Ventures. Its founding bet was to re-architect PostgreSQL with true separation of storage and compute — a custom Rust storage engine (Pageserver + Safekeeper with Paxos-like consensus) underneath a vanilla, unforked Postgres compute layer — so databases become serverless: autoscaling, scaling to zero, bottomless storage, free-tier-friendly. On top of that architecture Neon built capabilities no other Postgres service can match: instant copy-on-write branching (per-PR, per-developer, per-agent database environments), point-in-time restore ('time travel'), and live-migration autoscaling, all in a 100% PostgreSQL-compatible, Apache-2.0 open-source codebase (github.com/neondatabase/neon). Neon launched a technical preview in June 2022, opened signups in December 2022, reached GA on April 15, 2024 with 700,000+ databases managed, and grew to 'tens of millions of branches created every day' by 2026.
Neon owns the sharpest technical moat in serverless Postgres: it is the only open-source engine with a true storage/compute separation built from first principles (GetPage@LSN, safekeepers, pageservers), which yields structurally protected features — instantaneous branching, time travel, scale-to-zero — that competitors can't bolt on. That engineering gravity plus a developer-first free tier made it the default 'database for agents' before most of the industry noticed; at acquisition, 80%+ of Neon databases were created by AI agents. At ~$25M ARR and a ~$1B exit (roughly 40x, generous yet defensible for a category-defining infrastructure brand), the real strategic bet now sits inside Databricks: Neon becomes the OLTP layer of the lakehouse (Lakebase) and the 'boring backend' for agent apps, trading independence for distribution. The risk is commoditization — becoming an input to a platform just as Aurora, Azure Postgres, and Supabase attack the same wedge.
Browse: Home | Blog | About | All Startups A-Z | View full profile on StartupWiki