Apache Airflow has been the default data orchestrator for long enough to have become infrastructure in the boring sense, meaning it is everywhere, it is rarely chosen deliberately any more, and a very large number of data engineering jobs involve maintaining DAGs somebody else wrote in 2021. Astronomer is the commercial company most closely associated with the project and it runs the only certification programme that the market recognises. That combination makes Astronomer Academy more significant than its size suggests. The prep courses are free and they are properly about Airflow rather than about Astro, Astronomer's hosted platform.
That distinction matters and I checked it carefully, because the obvious commercial move would be to teach Airflow in a way that only makes sense on the vendor's runtime. The core material covers DAG authoring, the scheduler and executor model, task dependencies and trigger rules, XComs and the limits of passing data between tasks, sensors, connections and hooks, and the deferrable operator model. All of that is open source Airflow and works anywhere you run it. The split between Airflow 2 Fundamentals and Airflow 3 Fundamentals exams is a good decision.
Airflow 3 changed enough, particularly around the task execution interface and the scheduling model, that a single exam spanning both would have tested the wrong things. Maintaining two tracks costs Astronomer money and serves learners who are stuck on an older version, which is most of them, because upgrading a production orchestrator is the kind of project that gets deferred for years. Pricing is $150 per attempt, free for paying Astronomer customers, with student discounts available by emailing from a university address and periodic free vouchers distributed at events such as Airflow Summit. That is a reasonable structure.
It is not the cheapest credential I have seen, Starburst remains the benchmark at $49, but it is well inside the range where an individual can self-fund it, which is the test that matters for a credential aimed at working engineers. My reservations are editorial rather than commercial. The training does not engage with the case against Airflow, and by 2026 there is a real case. Airflow's model, where the DAG is a Python file parsed on a schedule and tasks are coordinated by a central scheduler, has known friction: local development and testing are awkward, passing data between tasks through XComs is a persistent source of bad design, dynamic workflows fight the parsing model, and the separation between orchestration and execution creates a class of bugs that only appear at scale.
Dagster, Prefect, Temporal and others exist because people found those limitations serious enough to build alternatives. A learner coming through this material will know Airflow thoroughly and will not know what the arguments against it are, which leaves them poorly equipped for the meeting where somebody proposes a migration. The exam design leans toward configuration recall. Knowing the default value of a scheduler parameter is less interesting than knowing when a task should be deferrable, and a certification that tested design judgement more heavily would be a stronger signal.
That is a common failing across this category and not specific to Astronomer, but it is worth noting since the credential's value depends on what it actually measures. The counterweight to all of this is straightforward. Airflow is named in an enormous number of data engineering job descriptions, this is the only certification anybody recognises for it, the training that prepares you is free, and the exam costs $150. That is a good deal by any reasonable measure.
Four point two. Free, focused on the open source project rather than the commercial product, fairly priced, genuinely recognised, and quiet only about whether you should be using the thing at all.