A Look Back at the First Cloud Native Geospatial Conference – CNG 2025

In April 2025, the Cloud Native Geospatial Conference (CNG) brought us together in the majestic mountains of Salt Lake, at Snowbird, Utah. A setting that deeply resonates with our home base in the heart of the French Alps.

As a CNG member and Overture Maps partner, I had the honor of presenting my talk titled "Leverage cloud native vector data in your webmapping application, simply." I shared innovative techniques for manipulating and analyzing geospatial data in GeoParquet format, creating dynamic maps based on Overture data.

Cloud Native Geo: Breaking Down Data Silos

CNG brings together forward-thinking organizations in geospatial data access and processing. The idea is simple yet powerful: read data at the source, at scale, using standard web protocols instead of duplicating it. The goal? Break down silos, promote interoperability, and enable large-scale, cloud-based streaming workflows.

Pioneers like CARTO, Planet, Radiant Earth, and Development Seed helped spark this movement with technologies like STAC and COG, laying the groundwork for the cloud-native paradigm. Today, with the launch of Overture Maps—a global dataset distributed via PMTiles and GeoParquet—new tech giants (Meta, Microsoft, Amazon) are joining the ecosystem. The community is expanding: Fused, CarbonPlan, and many others are driving innovation and strengthening the movement.

Standards, Innovation, and Speed

CNG is working to accelerate cloud-native adoption by bringing stakeholders together, setting standards, and fostering an open approach. All of this, with a distinctly American mindset: “Move fast, break things.” This makes for an interesting contrast with more institutional organizations like the OGC, which move more cautiously but aim to establish standards as long-lasting foundations. The challenge? Move quickly, but build wisely, so the whole ecosystem can rely on solid ground.

Key Takeaways from the Program

• Zarr was the star of the show—featured in nearly every presentation. It’s becoming a go-to format for cloud-native setups, especially for managing multidimensional data.
• The community is still exploring how best to consume vector data in the cloud with GeoParquet: What’s the right backend? DuckDB? A browser setup with GeoArrow? Which solutions fit which use cases? For Overture data, re-partitioning Parquet files seems crucial to optimize performance.
• STAC also held the spotlight, with a new Parquet encoding (stac-geoParquet) and the idea of a catalog for Overture’s Parquet partitions.
• Finally, there was strong representation from the AI world—machine learning for satellite imagery extraction, and large language models (LLMs) to aid in search and decision-making within multi-agent systems.