Earthmover Marketplace Adds Global Agricultural AI Data Stack

According to a recent LinkedIn post from Earthmover, Taylor Geospatial has released a new global agricultural field-boundary dataset, “Fields of the World,” on the Earthmover Data Marketplace. The post indicates this is the third Taylor Geospatial dataset on the platform and positions the combination as a complete Earth observation stack for global agricultural AI applications.

The company’s LinkedIn post highlights that earlier datasets include global Sentinel-2 median mosaics for planting and harvest seasons, delivered as cloud-optimized, analysis-ready imagery in Zarr V3 format. It also notes availability of the AlphaEarth Foundations dataset, which provides Google DeepMind satellite embeddings from 2017 to 2025 for downstream AI model development.

As shared in the post, the Fields of the World dataset offers 2024–2025 field boundary predictions worldwide in Zarr V3, GeoParquet, and PMTiles formats, aimed at seamless integration into existing data pipelines. The LinkedIn content credits researchers from Arizona State University, Washington University in St. Louis, Clark University, and Microsoft’s AI for Good Lab with developing a novel architecture and training data for global-scale field boundary inference.

The post suggests that Wherobots RasterFlow, Taylor Geospatial, and NASA Harvest contributed to enabling global inference while addressing challenges such as data diversity and model generalization across heterogeneous agricultural landscapes. It emphasizes that the resulting EO stack—imagery inputs, pretrained embeddings, and boundary predictions—is delivered in cloud-native formats designed for planetary-scale workloads.

For investors, the post indicates that Earthmover is positioning its marketplace as an infrastructure layer for AI models built on satellite and Earth observation data, with a specific focus on agriculture. This could enhance the company’s role in the geospatial and agtech value chain by attracting research institutions, enterprises, and AI developers seeking ready-to-use, scalable datasets.

If adoption grows, such a stack may support recurring data-access revenues, deepen ecosystem lock-in, and increase switching costs for users integrating Earthmover’s marketplace into production workflows. The collaboration with academic and industry partners, plus references to models like PRUE and benchmarks like FTW, also signals alignment with cutting-edge research, which could improve Earthmover’s competitive positioning in the emerging market for AI-ready EO data.