Open Data Infrastructure
The Open Table Format Wars, Explained
A buyer-focused explanation of Iceberg, Delta, and Hudi format competition, where convergence helps, and where lock-in still hides.
The open table format wars are not really about which logo wins. They are about where control lives when many engines need to read, write, govern, and repair the same data.
Why table formats became strategic
Data lakes needed more than files. They needed snapshots, schema evolution, deletes, partition evolution, time travel, and safe concurrent operations. Iceberg, Delta, and Hudi all respond to that need from different histories and ecosystems.
The competition matters because table formats define the contract between storage and compute. If the contract is open and broadly implemented, customers get more engine choice. If the contract is controlled by one platform in practice, the data may look open while the operating model stays captive.
Convergence is real, but uneven
The ecosystem is moving toward more interoperability. Delta UniForm and Apache XTable are examples of efforts to bridge format boundaries through metadata. That helps buyers because mixed estates are normal. Nobody wakes up with a perfectly clean lakehouse.
But convergence does not mean sameness. Feature support, write behavior, catalog integration, governance, and operational tooling still vary. The interesting question is not which format can read which files. It is which contract your production systems can trust.
The politics are about ecosystems
Format choices pull catalogs, engines, vendors, and communities along with them. Iceberg has strong neutral-governance appeal through the Apache Software Foundation. Delta has deep Databricks gravity and open-source roots. Hudi has a long history around incremental processing and lakehouse writes.
Core idea: the format war matters less than the control plane that forms around the winning patterns.
The buyer question
Buyers should not ask "which format is best?" in the abstract. Ask which format has the strongest ecosystem for your workloads, the clearest governance story, the fewest private assumptions, and the best path for the engines you need.
The winner is not the format with the loudest booth. It is the format that keeps your future architecture from getting smaller.
Sources to start with
These are the primary sources I would start from when checking the claims in this piece.