Earthmover – Weekly Recap

Earthmover featured prominently in climate and data infrastructure news this week, with multiple updates underscoring its role as an AI-ready platform for weather, ocean, and geospatial analytics. The company used a series of LinkedIn announcements to highlight marketplace expansion, new technical capabilities, and customer case studies.

Earthmover promoted an upcoming webinar with Spire focused on an AI-driven sub-seasonal-to-seasonal weather forecast model that delivers 46-day outlooks using data from a large commercial satellite constellation. The dataset is distributed via the Earthmover Data Marketplace, with content tailored to energy, insurance, and climate risk teams seeking extended-range forecasts.

In parallel, Earthmover showcased a case study with Amsterdam-based Beyond Weather, an AI forecasting startup serving energy and agri-food markets. Beyond Weather adopted Earthmover’s Flux API, Icechunk, and Arraylake to rebuild its visualization and data workflows, reportedly cutting load times to milliseconds and simplifying infrastructure maintenance while listing its wind forecast products on the marketplace.

The company also expanded its marketplace with a new oceanography segment, adding ocean model hindcasts as cloud-native data cubes alongside existing atmosphere and land offerings. Initial datasets from the U.K. National Oceanography Centre and Italy’s OGS are free, open, and optimized for machine learning, targeting users in climate and ocean research as well as commercial analytics.

On the open-source front, Earthmover highlighted the experimental Rectilinear Chunk Grid feature in Zarr-Python 3.2, enabling variable-length chunking for large, irregular arrays. The capability, expected to stabilize in Zarr v3.3, is designed to improve efficiency for scientific and geospatial archives such as ERA5, NetCDF, GRIB, and domain-partitioned datasets.

Another update detailed improvements in Icechunk V2, Earthmover’s tensor-data storage layer built for cloud object stores and high-concurrency environments. Stress testing with tools like AWS Labs’ Shuttle, Proptest, and Shopify’s Toxiproxy aims to harden consistency, fault tolerance, and network behavior for multi-terabyte, always-on workloads.

Commercially, Earthmover continued to broaden its AI-ready geospatial stack by adding Taylor Geospatial’s “Fields of the World” dataset, combining global field boundaries with Sentinel-2 mosaics and AlphaEarth embeddings for agricultural AI applications. The platform also highlighted prior work with Planette AI and Columbia Climate School, underscoring support for both near-term forecasts and long-range climate projections.

Taken together, this week’s developments illustrate Earthmover’s strategy of deepening its marketplace, strengthening core open-source infrastructure, and showcasing real-world customer integrations. These moves may enhance platform stickiness across climate, energy, agriculture, and insurance use cases and reinforce its positioning in AI-enabled environmental and geospatial data markets.