CedarDB opens limited access to the cloud version of its database that blends OLTP and OLAP

The problem CedarDB is attacking

Most engineering teams live with two separate databases. An operational database (OLTP) stores orders, payments and records in real time. An analytical database (OLAP) or data warehouse answers questions about trends in that data, usually fed by ETL pipelines that replicate and transform the information with a delay.

That separation works well when analytics can wait for an overnight report. It breaks down when an AI agent needs to answer, in seconds, a natural-language question about inventory or a specific customer's history by querying data that was just written. It was this mismatch that CedarDB co-founder and CEO Moritz Sichert described to TechCrunch as the reason behind the startup's recent growth.

From academic project to product

CedarDB is a spinout from the Technical University of Munich (TUM) based on Umbra, a research project built from scratch to outpace legacy database engines in processing speed by taking advantage of modern hardware. The company, headquartered in Munich, has five Umbra researchers as co-founders: Moritz Sichert, Lukas Vogel, Philipp Fent, Dominik Durner and Christian Winter.

In a post on the company's blog, Winter detailed the priorities the team set after leaving academia.

such as reliability, fault tolerance and compatibility with popular tools.

Christian Winter, co-founder of CedarDB

Postgres compatibility as strategy

According to Sichert, the team made sure CedarDB is compatible with PostgreSQL, an open source database system. For him, this choice keeps switching costs low, since users don't need to rewrite their application to migrate.

There's a second advantage, Sichert points out: PostgreSQL is the SQL dialect LLMs are most familiar with, because of the data used to train those models. That makes it easier for AI agents to generate correct queries against the database without needing a dialect translator in the middle.

For developers in Brazil, where Postgres is already the default choice across much of the stack (from startups to digital banks), this protocol compatibility means, in theory, that drivers, ORMs and administration tools already in your project keep working without rewrites. It's worth testing before trusting: protocol compatibility isn't the same as full feature and extension parity.

Enterprise first, cloud only now

CedarDB took a decade to get here, but its commercial structure is recent. In May 2025, the company announced, at the same time, a €5.3 million seed round (about US$6 million), CedarDB Enterprise (a paid, on-premise version) and CedarDB Community Edition (a free version).

The team debated internally whether it should launch a cloud offering from the start. It decided to delay, not abandon, the idea: after beta tests with users signed up on a waitlist, CedarDB announced this Wednesday the limited availability of CedarDB Cloud, on stage at TechCrunch Disrupt, where the startup is also competing in the Startup Battlefield against another 20 companies.

A fully managed cloud version tends to open up CedarDB's user base beyond large enterprises, which until now were the predominant customer profile for the on-premise version. It's also the move that puts the German startup face to face with larger competitors aiming at the same problem.

Databricks targets the same thesis

Databricks, a large data company, reached a similar conclusion. In June 2026, at the Data + AI Summit, the company presented LTAP, a new architecture that promises to unify transactions, analytics, streaming and operational data in a single storage copy inside the data lake.

The justification Databricks gave for the launch was explicit.

the agentic era

Databricks, at the June 2026 Data + AI Summit

Sichert reads Databricks' move as one more sign that there's real demand for what CedarDB offers, even though it's still early to know how much this unification will actually change the database market.

What's still open

CedarDB plans to serve its first cloud users with its current team of 14 people, but Sichert already anticipates the need for a bigger team once traction picks up.

And this is where we would raise our next round.

Moritz Sichert, CEO and co-founder of CedarDB

For those evaluating adopting CedarDB in production in Brazil, it's worth weighing that the cloud launch is in limited availability, not general availability: the operating track record, SLAs, the presence of a local data region and Portuguese-language support are all still unknowns.

The idea of combining OLTP and OLAP in a single engine isn't exclusive to CedarDB, a category that includes names like TiDB, SingleStore, CockroachDB and AlloyDB, each with different trade-offs between compatibility, maturity and consistency model. CedarDB's differentiator, for now, is its explicit bet on the PostgreSQL protocol as a bridge between legacy applications and the growing need for AI agents that query operational data in real time.