State of Airflow 2026: The Orchestration Layer is Uniting Data, AI, and Enterprise Growth
8 min read |
Every enterprise runs on data. But the enterprises winning today? They’ve realized that orchestration defines their competitive edge.
We’ve just released the State of Airflow 2026 report, built on insights from over 5,800 data professionals across 122 countries—the largest survey of data engineers ever conducted. The findings tell a story I’ve been watching unfold from my seat on the Apache Airflow Project Management Committee: Airflow isn’t just growing. It’s becoming the operational backbone that unites the tools, teams, and technologies powering intelligent data platforms worldwide.
The breadth of what Airflow now powers
What strikes me most about Airflow’s evolution isn’t just the scale, though the numbers are impressive. It’s the breadth of personas we now serve, each with fundamentally different jobs to be done.
We have data engineers building the pipelines that move mountains of data. Data analysts powering structured analysis that drives decisions. ML engineers creating domain specific models and AI engineers using foundation models to create applications that feel like magic. And increasingly, business experts using Airflow through human-in-the-loop mechanisms to ensure AI applications remain reliable and predictable.
This creates something profound: the potential for Airflow to become ubiquitous with context graphs, for cross-system context engineering to best understand not just the results of actions, but why those actions were taken.
Airflow 3 is accelerating everything
The release of Airflow 3 in April 2025 marked the project’s most transformative moment yet. It was the first major version in over four years and the largest release in Airflow’s history, bringing capabilities purpose-built for AI workloads, dramatic developer experience improvements, stronger security, and the flexibility to run tasks anywhere, at any time.
The adoption numbers tell the story: 26% of Airflow users have already upgraded to Airflow 3—remarkable for a major release less than a year old. Among large enterprises with 50,000+ employees, where upgrade decisions move slowly and requirements are stringent, nearly 20% are already running Airflow 3.
For our Astro customers, the numbers are even more striking. 48% are already on Airflow 3, including an incredible 60% of our largest enterprise customers. This early adoption at scale demonstrates deep enterprise confidence in both the Airflow project and the value of new capabilities.
Among those still on 2.x versions, 84% are already planning their upgrade. Features like event-driven scheduling, Dag bundles, and remote execution aren’t nice-to-haves anymore. They’re essential capabilities for modern data operations.
AI is moving from prototype to production on Airflow
This won’t come as a surprise, but one of the major drivers for the timing of the Airflow 3 release was generative AI. The rise of GenAI created an urgent need for orchestration that could support far more complex use cases than traditional data pipelines. With architectural changes and features purpose-built for AI workloads, Airflow 3 enables organizations to move from prototypes in notebooks to production-grade generative AI initiatives.
The data confirms this shift is happening: 32% of Airflow users now have GenAI or MLOps use cases in production, a five percentage point increase year-over-year.
Among Astro customers, that number doubles to 62%. For our most mature customers who’ve been with us since 2022, it climbs to 83%. These aren’t experiments. These are production AI systems that generate revenue, serve customers, and drive business outcomes.
Companies like Laurel, which raised $100M for the world’s first AI timekeeping platform, use Astro to orchestrate the complex workflows powering their rapidly growing business. This includes RAG pipelines, LLM context engineering, and inference workflows that must work reliably at enterprise scale.
But here’s what makes this even more compelling: the leaders of the AI movement itself run on Airflow. At the 2025 Airflow Summit, OpenAI, Anthropic, GitHub, and countless other leading organizations presented on how Airflow is critical to their businesses. OpenAI standardized on Airflow in 2023 and now runs ~7,000 pipelines with near-universal usage across the company. GitHub’s Copilot—deployed by 90% of the Fortune 100—depends on Airflow pipelines to continuously aggregate engagement metrics, quality indicators, and user feedback.
When the companies building the future of AI depend on Airflow to power their own operations, that tells you something.
Airflow drives revenue, not just dashboards
Here’s a trend that would have seemed unlikely five years ago: 89% of users expect to use Airflow for more revenue-generating or external-facing solutions next year.
Orchestration used to be about internal analytics and operational efficiency. That’s changing rapidly. Airflow now powers customer-facing applications, AI-driven products, and business processes that directly impact the bottom line. Airflow is not just moving data - it’s enabling business growth.
This shift is having a real impact on careers too. 94% of respondents say having a strong foundation in Airflow skills will positively impact their career over the next five years, with over a third saying Airflow is central to their job and career trajectory.
The messy reality: Airflow tames heterogeneous data stacks
Enterprise technology is messy. Organizations rarely use a single platform. Strategy shifts, leadership changes, mergers and acquisitions, and best-of-breed approaches all contribute to heterogeneous data stacks. The only constant is the need for a flexible, agnostic orchestrator to tame it all.
Airflow has established itself as exactly that foundation.
Our survey data on which tools Airflow users pair with reveals fascinating insights into the data landscape: Snowflake and Databricks remain neck and neck for adoption among Airflow users, while more than 22% leverage two or more of the three most popular cloud-based data platforms. This isn’t platform lock-in—it’s platform flexibility.
The most commonly paired tool with Airflow for the second year running? dbt, with 44% adoption. Astronomer’s open-source Cosmos package, which helps users integrate dbt and Airflow, saw more than 200 million downloads in 2025.
Your data platform needs a foundation that lets you plug and play the best tools for your business without locking you in or requiring months of reworking pipelines. That foundation is Airflow.
Mission-critical means different things at scale
As organizations scale their data operations and their stacks become more complex, Airflow becomes increasingly important. Over half of respondents at large enterprises said Airflow is very important to their company.
Companies like Autodesk have turned to managed Airflow to modernize and stabilize the pipelines powering their most essential analytics and product intelligence workflows. As Nick Wilson, Senior Manager for Autodesk’s Platform team, shared: “From the standpoint of stability and of being on the bleeding edge of what you can do with Airflow, Astronomer is far ahead of any other provider.”
Remote execution functionality in Airflow 3 directly supports running sensitive workloads for highly regulated industries—data subject to GDPR in eCommerce, HIPAA in healthcare, mission-critical data that can never leave on-prem servers in banking and financial services.
Addressing the productivity paradox
Here’s an irony in the AI movement: data engineers aren’t yet benefiting from the productivity gains AI can deliver. Generic AI tools are accessible but problematic for specific data engineering work. 43% of users cite hallucinations as a major issue, 42% struggle with outdated syntax, and only 9% are happy with Dags generated by generic AI tools.
This is why we built Astro IDE, which is an in-browser Dag writing experience with an Airflow-specialized AI assistant, designed to 10x data engineer productivity. Early feedback has been remarkable. As Tim Handley, Chief Product & Technology Officer at Welbee, told us: “Astro IDE generated code that was high-quality and production-ready from the start. Compared to generic AI tools, Astro IDE was miles ahead.”
What makes me optimistic
There are many milestones in Airflow’s growth trajectory. 43,800+ GitHub stars, 3,600+ contributors, more than 300 individuals contributing to Airflow 3.0 alone. But what makes me most optimistic isn’t the metrics.
It’s seeing Airflow evolve from a workflow orchestrator into something far more significant: the operational platform unifying data, models, and decisions. The foundation for enterprises where humans and AI agents work side by side. The orchestration layer that makes heterogeneous, complex data ecosystems not just manageable, but powerful.
Airflow 3 was the biggest milestone yet, and it’s just the beginning. As you’re reading this, development is underway on future releases informed by this survey and feedback from thousands of production deployments.
88% of survey respondents would recommend Airflow to others. This shows confidence in a platform that’s powering AI applications, mission-critical business processes, and revenue-generating data products at scale.
All signals suggest that Airflow’s biggest days lie ahead of us.
Download the complete State of Airflow 2026 report for deeper insights, real-world case studies, and detailed analysis of how industry leaders are scaling Airflow to drive AI, automation, and business transformation.
Ready to unlock the full potential of Airflow for your organization? Let’s talk.
