wouldit.work

Developer tools

Neon: serverless Postgres sold to Databricks for about $1B as AI agents took over

Neon built a serverless Postgres that spins up a database in under a second. Four years after its founding, Databricks agreed to buy it for about $1 billion, citing telemetry that more than 80% of new Neon databases were created by AI agents.

Product
Neon
Founder
Nikita Shamgunov, Heikki Linnakangas and Stas Kelvich
Launched
2021
Model
Free tier plus usage-based paid plans

Published · 2 min read · 2 sources

Key takeaways

  • Neon was founded in 2021 by Nikita Shamgunov, Heikki Linnakangas and Stas Kelvich, and raised $129.6 million according to Crunchbase data cited by TechCrunch.
  • On 14 May 2025 Databricks announced it would acquire Neon; TechCrunch reported the price as about $1 billion.
  • Databricks said internal telemetry showed more than 80% of databases provisioned on Neon were created automatically by AI agents rather than by humans.

What it is

Neon is a managed, serverless Postgres database. Compute and storage are separated, it scales automatically, and developers can branch a database (schema and data) to test changes before production (TechCrunch). It offers free and usage-based paid plans (TechCrunch). Databricks says Neon can spin up a fully isolated Postgres instance in 500 milliseconds or less (Databricks press release).

Timeline

  • 2021: Neon is founded by CEO Nikita Shamgunov and software engineers Heikki Linnakangas and Stas Kelvich (TechCrunch), "a team of experienced database engineers and Postgres contributors" (Databricks press release).
  • By May 2025: Neon has raised $129.6 million according to Crunchbase, from investors including M12, General Catalyst, Menlo Ventures and Notable Capital (TechCrunch).
  • 14 May 2025: Databricks announces its intent to acquire Neon, saying "over 80 percent of the databases provisioned on Neon were created automatically by AI agents rather than by humans" (Databricks press release). TechCrunch reports the deal at about $1 billion (TechCrunch).

Why it worked

  • Its design fit a new kind of user. Databricks lists three things agents need: provisioning at machine speed, cost that scales with usage, and plain Postgres compatibility. Neon's instant databases and branching matched all three (Databricks press release).
  • Cheap, disposable databases. Separating compute from storage keeps the cost of "thousands of ephemeral databases proportional to the queries they actually run" (Databricks press release).
  • Deep Postgres expertise. The founding team were Postgres contributors, and Neon is "100 percent Postgres-compatible" (Databricks press release).

What to learn

  1. Check who is actually signing up. Ali Ghodsi summed up the deal as "four out of every five databases on their platform are spun up by code, not humans" (Databricks press release).
  2. A free plan plus instant, API-driven setup makes it easy for other software to adopt you without a sales call; Neon offered free and usage-based plans (TechCrunch).
  3. Compare Supabase, which reports the same AI-driven pattern on its own Postgres platform, and PlanetScale, which went the other way and dropped its free tier.

Risks and what didn't work

  • The headline numbers are not audited. The 80% figure is the buyer's own "internal telemetry", and the roughly $1 billion price is TechCrunch's report, not a disclosed term (Databricks press release; TechCrunch).
  • Agent-driven usage can be shallow. Many agent-created databases are ephemeral by design; the press release itself frames them as "thousands of ephemeral databases" rather than long-lived paying customers (Databricks press release).
  • Independence ended. The team was expected to join Databricks after closing, which was subject to regulatory clearances (Databricks press release).

FAQ

Short answers drawn from the sections above.

What is Neon?

Neon is a managed, serverless Postgres database. Compute and storage are separated, it scales automatically, and developers can branch a database (schema and data) to test changes before production (TechCrunch). It offers free and usage-based paid plans (TechCrunch). Databricks says Neon can spin up a fully isolated Postgres instance in 500 milliseconds or less (Databricks press release).

Why did Neon work?

  • Its design fit a new kind of user. Databricks lists three things agents need: provisioning at machine speed, cost that scales with usage, and plain Postgres compatibility. Neon's instant databases and branching matched all three (Databricks press release).
  • Cheap, disposable databases. Separating compute from storage keeps the cost of "thousands of ephemeral databases proportional to the queries they actually run" (Databricks press release).
  • Deep Postgres expertise. The founding team were Postgres contributors, and Neon is "100 percent Postgres-compatible" (Databricks press release).

What can indie builders learn from Neon?

  1. Check who is actually signing up. Ali Ghodsi summed up the deal as "four out of every five databases on their platform are spun up by code, not humans" (Databricks press release).
  2. A free plan plus instant, API-driven setup makes it easy for other software to adopt you without a sales call; Neon offered free and usage-based plans (TechCrunch).
  3. Compare Supabase, which reports the same AI-driven pattern on its own Postgres platform, and PlanetScale, which went the other way and dropped its free tier.

What didn't work for Neon, and what are the risks?

  • The headline numbers are not audited. The 80% figure is the buyer's own "internal telemetry", and the roughly $1 billion price is TechCrunch's report, not a disclosed term (Databricks press release; TechCrunch).
  • Agent-driven usage can be shallow. Many agent-created databases are ephemeral by design; the press release itself frames them as "thousands of ephemeral databases" rather than long-lived paying customers (Databricks press release).
  • Independence ended. The team was expected to join Databricks after closing, which was subject to regulatory clearances (Databricks press release).

Sources

Every link cited above, in order of first use.

  1. 1.TechCrunch (techcrunch.com)
  2. 2.Databricks press release (databricks.com)

Related case studies

More in Developer tools →