— Stack · Integrations

dbt on AWS

The transformation layer of a data platform. Your SQL transformations become versioned, tested code that runs on the engine you already use. dbt Core is open source; the dbt platform adds a hosted scheduler, IDE and CI for a licence.

How it fits

dbt builds the layers between raw and ready. Your engine does the compute.

Support levels from docs.getdbt.com: v2 adapter lifecycle table, v1 pages for Athena (maintained by dbt Labs) and Glue (community, not supported by dbt Labs). Ingestion into bronze is done by other tools; dbt starts from tables already landed.

What dbt is doing

SQL, run like software.

Versioned

Stored procedures, retired

Each transformation is a SELECT in a file, reviewed and merged in Git like application code. It replaces stored procedures and the bash scripts that chain SQL files in order.

Modular

Write a rule once

Models build on each other, so a business rule lives in one model and every downstream table reuses it. dbt works out the build order from those links.

Environments

Same code, dev to prod

One project builds dev, staging and production. Only the target changes, so what you tested is what runs in production.

Trust

Tested before it ships

Tests and docs sit next to the SQL. Each pull request builds and tests the changed models in a temporary schema before the merge.

The three choices that matter

Who runs it, on what, for how much.

Who triggers the dbt run

dbt Core is a command-line tool, so something has to run it on a schedule and keep the logs and artefacts. Pick the orchestrator your team already operates.

Core or the dbt platform

dbt Core is free and runs anywhere you schedule it. The dbt platform adds a scheduler, hosted IDE, lineage catalog and CI, billed per seat plus a cap on models built.

What drives the cost

A dbt run is a series of engine queries. The warehouse or query engine is where most spend lands, so model SQL and incremental builds are the main cost lever.

Dot counts are illustrative, not measured; the split depends on your engine and run frequency.

What you pay

A dbt-on-AWS bill, in two parts.

The engine

runs every model

Athena at $5 per TB scanned, or Redshift Serverless from $1.50 per hour; this is where most spend lands

dbt Core

free

Open source; you pay only to host and schedule it (MWAA, ECS, Step Functions)

dbt Developer

free

1 seat, 3,000 successful models built per month, 1 project

dbt Starter+

seats + licence

Starter $100 per user per month; Enterprise custom, both metered on successful models built

What we see in production

One repository for every team, or one per team with copied models.Decide the split early: one project per domain, with shared models consumed through packages or cross-project references.

Hundreds of models, no owner, no layers.Name models by layer, give each one an owner, and delete the ones nothing downstream reads.

Every dashboard computes revenue its own way.Define metrics once in a semantic layer so BI tools and agents read the same number.

Full-refresh models rebuilding the whole table nightly.Make large models incremental so each run only processes new rows.

Athena models that scan the entire dataset per run.Partition and filter; Athena bills $5 per TB scanned every build.

A warehouse left running for a job that takes minutes.On Redshift Serverless or Snowflake, let compute pause between dbt runs.

How we help

Industrial-grade data pipelines

Use cases in production, not POCs: tests, CI/CD, lineage, on-call that sleeps at night.

Read the approach →

PDF · coming soon

The one-page dbt on AWS field guide.

Best practices, cost traps and performance checks, in the order you should run them. Printable, no fluff.