When competing against a cloud hyperscaler, a database is an excellent place to start. Low churn (data gravity), expensive products (often not fully utilized), and naturally built-in net dollar retention (databases rarely shrink). A look at two separate but converging spaces of database-as-a-service (DBaaS) and backend-as-a-service (BaaS).

DBaaS is what it sounds like – e.g., vanilla or specialized managed Postgres or MySQL. BaaS extends the product offering – usually with building blocks like authentication, authorization, and API gateways in addition to a managed DBaaS.

Pure play DBaaS are hard to build nowadays. The successful ones (e.g., MongoDB, Neo4j, Snowflake, ClickHouse1) all started over a decade ago and benefited from cloud adoption tailwinds. Today, hyperscalers offer competitive alternatives, making it difficult for newcomers to differentiate.

It follows that DBaaS have started to move up the stack.

For example, take a look at serverless databases like PlanetScale or Neon. While the technology is differentiated, the most considerable appeal for most developers is a pristine developer experience. For example, PlanetScale lists its developer CLI as the second link in its navigation bar after documentation. While Vitess (the technology behind PlanetScale) has users at a significant scale, I'd imagine many users are looking for a serverless database for their frontend framework, e.g., Netlify or Vercel.

Snowflake is also moving up the stack towards BaaS/platform. Their $800m acquisition of Streamlit provides a data-science notebook experience for users to explore and visualize data. Recently, they've launched an application platform to power applications on top of the data warehouse.

Compare this to Supabase, a BaaS, which has taken a similar route in OLTP. They held the managed database constant (Postgres) and built the application platform around it (etc., authentication, API gateway).

Why does it matter?

There are two questions in my mind:

  1. Is this natural product expansion or pressure from the underlying cloud hyperscalers?
  2. Does value accrue on DBaaS or BaaS end of the spectrum?

BaaS benefit from greater lock-in and data gravity and a more direct relationship with the end-user – they can't as easily be abstracted away because they offer a much larger API surface.

On the other hand, DBaaS scale better with larger customers (ultimately, these are all enterprise businesses) than BaaS, which is a small step away from the fragmented PaaS space (see Why Did Heroku Fail?).

1ClickHouse the company is less than a year old, but the technology is over ten years old, initially built at Yandex.