Earthmover: Cloud-Native Scientific Data Platform

Earthmover is a cloud-native data platform for n-dimensional array data, or tensors. It was purpose-built for Earth System and climate data, designed to make massive environmental and geospatial datasets analysis-ready, versioned, and queryable directly in your S3-compatible cloud storage.

Founded by climate scientists and open-source maintainers of Zarr, Xarray, and Icechunk, Earthmover transforms legacy file-based workflows into scalable, AI/ML-ready data cubes optimized for performance, governance, and collaboration.

A governed data catalog provides unified discovery, lineage tracking, and fine-grained access control across teams and tools. Data stored in Arraylake streams seamlessly to models, notebooks, GIS applications, and GPUs, enabling reproducible pipelines, rapid scenario testing, and transparent versioning.

Scientists and analysts can build upon each other's work without duplicating effort or losing context, accelerating analysis while maintaining full traceability.

For energy and commodities trading teams, the payoff is in the velocity of modeling execution. Weather is one of the most powerful drivers of market volatility, yet most data teams spend more time preparing data than forecasting.

With Earthmover, datasets are pre-structured for computation, allowing models to run faster, forecasts to update continuously, and strategies to respond in real time. Fast query times saturate GPUS with physical AI data that maximize precious compute resources.

Teams cut time-to-model from weeks to minutes and reduce storage and compute costs, turning complex physical-world data into actionable, timely trading intelligence.