The lakehouse has quietly crossed a tipping point. Every major cloud provider now speaks the same language: Iceberg REST Catalog. AWS S3 Tables, Google BigLake, Cloudflare R2, Snowflake Polaris, Databricks Unity Catalog, and Microsoft OneLake all offer native Iceberg support with REST catalog interfaces. For the first time, a single query engine can read and write tables across all of them without vendor-specific connectors. But open formats are only half the story. A new generation of lightweight, native query engines - including DataFusion, DuckDB, and LakeSail - can now run directly against these open tables with minimal infrastructure. No clusters to provision, no heavy runtimes to manage. Just fast, efficient compute on open data, whether you are running locally, in a notebook, or embedded in an application. In this session, I will share a demonstration of a single DuckDB process connecting to five Iceberg REST catalogs across five different cloud providers, reading and writing real data with a unified SQL interface. I will walk through the practical evolution of the native lakehouse stack, from delta-rs reading Delta tables to DataFusion powering query execution and show how these engines are reshaping what "lakehouse" means in practice. Finally, I will show how AI coding assistants are accelerating this entire journey; helping engineers write lakehouse queries, build data pipelines, and navigate the rapidly evolving ecosystem of open table formats and catalog APIs faster than ever before. Whether you are evaluating lakehouse architectures, building multi-cloud data platforms, or simply curious about where the industry is heading, this talk will give you a practical, vendor-neutral view of the native lakehouse in 2026.
The lakehouse has quietly crossed a tipping point. Every major cloud provider now speaks the same language: Iceberg REST Catalog. AWS S3 Tables, Google BigLake, Cloudflare R2, Snowflake Polaris, Databricks Unity Catalog, and Microsoft OneLake all offer native Iceberg support with REST catalog interfaces. For the first time, a single query engine can read and write tables across all of them without vendor-specific connectors. But open formats are only half the story. A new generation of lightweight, native query engines - including DataFusion, DuckDB, and LakeSail - can now run directly against these open tables with minimal infrastructure. No clusters to provision, no heavy runtimes to manage. Just fast, efficient compute on open data, whether you are running locally, in a notebook, or embedded in an application. In this session, I will share a demonstration of a single DuckDB process connecting to five Iceberg REST catalogs across five different cloud providers, reading and writing real data with a unified SQL interface. I will walk through the practical evolution of the native lakehouse stack, from delta-rs reading Delta tables to DataFusion powering query execution and show how these engines are reshaping what "lakehouse" means in practice. Finally, I will show how AI coding assistants are accelerating this entire journey; helping engineers write lakehouse queries, build data pipelines, and navigate the rapidly evolving ecosystem of open table formats and catalog APIs faster than ever before. Whether you are evaluating lakehouse architectures, building multi-cloud data platforms, or simply curious about where the industry is heading, this talk will give you a practical, vendor-neutral view of the native lakehouse in 2026.
Project Manager
Mimoune "Mim" Djouallah, formerly a Construction Planner, joined the Microsoft Fabric CAT team in December 2023. He holds a BSc in Civil Engineering and has worked extensively with the Power BI stack since 2016. As an early adopter of Microsoft Fabric, Mim shares hands-on insights through blogging, focusing on Python notebooks, GIS, and data engines that integrate with OneLake.
