CesiumJS is building on our core 3D Tiles capabilities to support a broader range of tiled vector data. This talk is a technical deep dive into that work:

• How CesiumJS is approaching Mapbox Vector Tiles, MapLibre-compatible sources, and vector data from Architecture, Engineering, and Construction (AEC) workflows
• How those capabilities fit alongside the 3D Tiles open standard

These data sources come from ecosystems with very different assumptions about projection, tiling, styling, feature identity, precision, and level of detail. Supporting them in a single 3D geospatial engine alongside 3D reality data such as meshes, imagery, point clouds, and Gaussian splats requires identifying which concepts can be shared across formats and which remain specific to their source ecosystems.

We will walk through the vector rendering pipeline from data source to pixels: discovering and requesting tiled content, decoding geometry and attributes, transforming coordinates, generating GPU-ready representations, styling features, and managing resources as the camera moves.

We will discuss the challenges of adapting a Web Mercator, 2D map-oriented model to a 3D globe, including reprojection, tile-edge behavior, zoom-level selection, draping, line and polygon rendering, and style interoperability.

Finally, we will compare key aspects of different vector approaches with 3D Tiles, including spatial subdivision, level of detail, metadata, streaming, and feature selection. Rather than treating vector formats and 3D Tiles as competing solutions, we will show how they can coexist within the same application and where shared runtime infrastructure becomes valuable.

Attendees will leave with a clearer understanding of the emerging vector architecture in CesiumJS, the tradeoffs involved in unifying these workflows in a 3D-first engine, and how that work relates to the broader evolution of 3D Tiles and heterogeneous 2D/3D data on the web.