Serverless Postgres

Almost every app and every website needs to remember things: user accounts, orders, comments. For this you use a database, a program that stores such data in an organized way and hands it back out on request. PostgreSQL, or Postgres for short, is one of the world’s most widely used programs of this kind and is available for free. Normally it runs on a computer in a data center that stays switched on around the clock and costs money around the clock. Serverless Postgres is the same database, but as a rental service that adjusts itself to demand. When many requests come in, it automatically gets more computing power. When not a single one comes in for days, it goes to sleep and costs almost nothing.

Why developers no longer want to rent a server

The name is a bit misleading: servers still exist, of course. What is meant is that nobody has to take care of them anymore. Anyone who runs a database the classic way chooses a server size, installs updates, sets up backups, and monitors utilization. With a serverless service, the provider takes care of all of that completely. For small teams, this is often the difference between having your own technician and having none.

The second reason is money. A rented database server costs money even at night, on weekends, and during the testing phase when nobody is using the app yet. This is exactly where startups burn a lot of budget on idle time. With Serverless Postgres, you pay roughly by compute-seconds and gigabytes stored. A project that only really works a few hours per month can drop from 50 euros to just a few euros this way.

The third point is important: load spikes. If an app is suddenly shared on social media, a fixed-size server will eventually collapse. A serverless service scales up within seconds and then back down again afterward. So you don’t have to guess the peak in advance and pay for it all year round.

The separation of computing and storage

The decisive technical trick is a division. In a classic Postgres installation, the data sits on the hard drive of exactly the machine that also processes the requests. The two are tightly bound together. Serverless providers separate these two tasks: the data lives in its own storage system on the network, which is always available. The computing work is done by a separate process that can be started, scaled up, or shut down at any time.

You can picture this like a warehouse with a changing workforce. The shelves with the goods stand there permanently. How many people are packing parcels at the same time depends on the order volume. If no orders come in, the staff goes home without the goods disappearing. That’s exactly why the computing power can drop to zero without any data being lost.

This design has a price you should be aware of. If the database has been asleep, it first has to boot back up on the next access. This delay is called a cold start and, depending on the provider, lasts anywhere from milliseconds to a few seconds. In addition, there is now a network between computing and storage, which can slow down individual queries slightly. For applications with a steady, high, continuous load, a fixed server is therefore often still cheaper and faster.

Providers, price tags, and company acquisitions

Well-known names in this market are Neon, Supabase, Amazon Aurora Serverless, and Google AlloyDB. They all speak the language of normal Postgres to the outside world. That means existing program code usually works without modification, which makes switching easy. That is exactly their biggest selling point compared to entirely proprietary database systems.

In business news, the term mainly comes up in connection with funding rounds and acquisitions. Databricks bought the provider Neon in 2025 for around one billion dollars. The reason behind this is the AI boom: when programs automatically write code, they also create databases en masse for short tests. Such throwaway databases are only worthwhile if they can be created in seconds and cost nothing while idle.

You will also encounter the term with pretty much every smaller web project. Anyone building a club website, a school app, or a hobby project today gets a serverless Postgres database thrown in on the free tier with most hosting services. This should not be confused with Serverless Functions, where individual snippets of code run on demand instead of a database. The principle is the same, the product is different.