What are you actually paying a managed database provider for?
Not the database — PostgreSQL is free. You are paying for the parts that only matter when they fail: automated backups that have been tested, failover that works at three in the morning, version upgrades that do not lose data, and someone whose job it is when replication breaks.
That framing decides whether the premium is worth it. A single Postgres instance behind an internal tool does not need any of it, and self-managing on a €5 VPS is the right call. A production database whose loss would end the business needs all of it, and the difference between a managed service and a self-managed one at that point is not cost but survival.
Aiven prices from about $20 a month with a free tier for small services and usage-based billing beyond, which is more expensive than a self-managed instance and considerably cheaper than the equivalent AWS RDS and MSK combination. Scaleway, Clever Cloud and Exoscale all offer managed databases alongside their compute.
The question worth asking a provider is not what the SLA says but when the last restore test was run. A backup nobody has restored is a hypothesis.
Why does running standard engines matter more than any feature?
Because it determines what leaving costs. Every service in this category runs the actual open-source engine — Aiven runs real PostgreSQL, Kafka and ClickHouse; Nhost runs standard Postgres; PocketBase runs SQLite. Moving away means a dump and a restore, not a rewrite.
Contrast that with a managed service built on a proprietary fork or a vendor-specific query layer, where the application has been written against something only that vendor sells. The switching cost is then measured in engineering quarters rather than in hours, and the pricing conversation reflects it.
This is also why the European providers can compete without matching hyperscaler R&D budgets. They are not selling a better database; they are selling the same database, operated well, in a jurisdiction you chose — and the engine being identical is precisely what makes that a fair comparison rather than a compromise.
The practical check before committing: write down how you would move to another provider, and make sure the answer is a command rather than a project.
Managed service or managed-in-your-own-account?
There is a middle option most buyers do not know exists, and for regulated organisations it is the one that resolves the deadlock.
A normal managed service runs on the vendor's infrastructure under the vendor's account. Elestio inverts this: it deploys and fully manages more than 350 open-source applications inside your own cloud account or on dedicated European hardware, handling updates, backups, monitoring and TLS while the infrastructure — and the data — remain yours. Aiven offers the same shape for data services through Bring Your Own Cloud.
That matters when a compliance requirement says the data must sit in an account the organisation controls, which no conventional SaaS can satisfy. It also caps the exit risk: if the management layer disappears, the running systems are still in your account rather than somewhere you need to be given access to.
The trade is that you now pay two bills — the management fee plus the underlying infrastructure — and that the management layer itself is proprietary in both cases even though everything it deploys is open source.
When is a backend-as-a-service the right shape at all?
When the application needs a database, authentication, file storage and realtime updates, and you would otherwise assemble four services and the glue between them. That is the Firebase and Supabase model, and Europe has two credible answers.
Nhost pairs PostgreSQL with an instant GraphQL API, authentication with OAuth providers, file storage with image transforms and serverless functions, hosted in EU regions from a Swedish company — and the whole stack is MIT-licensed and self-hostable via Docker, which is a stronger position than most BaaS platforms offer. The trade is that GraphQL and Hasura add a learning curve, and the ecosystem is much smaller than Supabase's.
PocketBase takes the opposite approach: one executable file containing SQLite, auth, storage, realtime subscriptions and an admin dashboard. Deployment is copying a binary to a server, backup is copying a file, and there is no vendor at all. It is MIT-licensed and extensible as a Go framework.
The limit on PocketBase is architectural rather than a matter of maturity: SQLite has a single writer and scales vertically, so it fits small and medium applications, internal tools and prototypes, and does not fit a write-heavy system that needs to scale horizontally.
What does European jurisdiction change for a database?
It removes the CLOUD Act exposure, which for a database is more consequential than for most software because the database is where the personal data actually lives. A US-established provider can be compelled to produce data it holds regardless of which region the servers are in.
Aiven pins services to EU-only regions across several clouds while remaining a Finnish company. Scaleway keeps every data centre inside the EU in Paris, Amsterdam and Warsaw. Clever Cloud runs EU data centres from Nantes. Nhost offers EU regions from Stockholm. Elestio is Romanian and deploys into the region you choose.
Exoscale is Swiss, which is outside the EU with an adequacy decision covering transfers — lawful without standard contractual clauses, and not the same as intra-EEA processing for procurement rules that distinguish them. Its Austrian and Bulgarian regions are useful for latency in central and eastern Europe.
PocketBase sidesteps the question entirely: the binary runs wherever you put it, so there is no processor and no jurisdiction beyond your own hosting choice.
