Make smarter decisions on the cloud database market

The cloud database market is hot again, and AI apps are the reason.

We've found a lot of interesting signals. For example, ClickHouse raised $400M at a $15B valuation in January 2026, and bought Langfuse, an open-source tool for monitoring AI apps, in the same announcement. In June, Supabase raised $500M at $10.5B, about double its valuation eight months earlier. Investors are betting that AI apps and agents will need far more databases, and they're paying up for the ones developers already like.

In this section, we explain why AI has made databases exciting again, which kinds are winning the new workloads, and what these rounds mean for the databases the big clouds sell themselves.

Every app needs a database, and most teams end up running more of them than they can manage.

We compare six kinds of cloud database companies, from managed relational and document databases to vector, time-series and analytical engines. Their buyers range from solo developers and startups to large enterprises moving off old systems. Then we rank what they need most: lower bills, less time spent running databases, and enough speed when traffic spikes.

This shows which problems buyers will pay to fix now, and which ones they simply live with.

Usage-based fees, provisioned capacity, free tiers that turn into paid plans, enterprise contracts and sales through the big clouds' marketplaces: cloud database companies make money in many ways.

We compare who pays and what it costs to serve them. Most independent databases run on AWS, Azure or Google Cloud, so part of every dollar goes back to the cloud they compete with. Free tiers bring in millions of developers, but only a few turn into big accounts. And usage pricing grows with customers, but it also shrinks when they tune their queries.

You'll see which models keep healthy margins after paying the cloud bill, and which ones depend on customers never shopping around.

Any growth forecast is only as good as the reasons behind it. So here we lay out what actually pushes cloud databases forward.

Some drivers are plain to see: more apps, more data, and companies leaving the servers they run themselves. Others build on each other. AI apps need vector search and memory for their agents. Coding tools now set up a database with every new project. And the more data a company keeps in one database, the harder it is to move away.

Before building or investing, you should know what makes a team pick a database for a new project, and what keeps it there years later.

Snowflake and MongoDB are the names most people know in cloud databases, but they're only part of the story.

Plenty of strong teams work further from the headlines. Some build databases that agents can create and throw away in seconds. Others handle time-series data for factories and power grids, store graphs for fraud detection, or keep small databases running inside apps that sync to the cloud.

They're hard to compare because many are built on the same open-source engines, like Postgres. Two products can look alike on a pricing page and behave very differently under heavy traffic. And many sell through the big clouds' marketplaces, so outsiders rarely see how much they really sell.

Below, you'll find the market grids and ecosystem maps we've built to show which database does what, and where the newer engines fit.

Neon, PlanetScale, Pinecone… a few cloud database startups have won something real. They just haven't all won the same thing.

Neon made Postgres quick to start and cheap to pause, and Databricks bought it for about $1B as AI agents began creating databases. PlanetScale built its name running MySQL for very large apps, then added Postgres. Pinecone turned vector search into a product people paid for early, and now competes with databases that added vectors themselves.

We'll look at what these teams got right with developers first, and what it took to keep customers once bigger rivals offered something similar.