Earthmover

High-Level Overview

Earthmover is a cloud-native data platform designed for managing tensor, geospatial, weather, and climate data at exabyte scale. It builds products like Icechunk (cloud-optimized array storage), Arraylake (data management layer), and Flux (performant APIs) to store, collaborate on, and analyze multidimensional arrays (tensors) efficiently, outperforming traditional table-based systems.[1][4][5] Serving organizations in climate tech, renewables, insurance, research (e.g., NASA, University of Wisconsin-Madison), and AI/ML, Earthmover solves I/O bottlenecks, high storage costs, and collaboration challenges for massive scientific datasets, enabling faster R&D, model training, and operations—claiming industry-leading speed and cost savings.[1][3][4][5] With over 10 paying customers like Sylvera, Kettle, and RWE, and a $7.2M seed round in September 2025 led by Lowercarbon Capital, the company shows strong early momentum toward becoming the "Snowflake of weather and geospatial data."[1][5]

Origin Story

Earthmover was founded by Ryan Abernathey and Joe Hamman, who met collaborating on Xarray—a foundational open-source package for weather, climate, and geospatial analytics downloaded over 6 million times and used by NASA, ESA, Google, NVIDIA, and Planet.[1] Abernathey, transitioning from oceanography research amid the Pangeo project's growth (a 2017 grassroots effort now with 1,000+ participants powering tools like Zarr, Xarray, and Dask at NASA, Microsoft, and Google), spotted the need for a "modern data stack" tailored to scientific tensors rather than tables.[2] Hamman, a Pangeo collaborator and CarbonPlan founder, joined as CTO co-founder, leveraging their leadership in Zarr and the new Icechunk project.[1][2] The idea emerged from frustrations with "unstructured" data handling in business tools, pivoting open-source innovations into a commercial platform for climate and deep sciences, with early traction via case studies like Sylvera's carbon intelligence scaling.[2][4][5]

Core Differentiators

• Tensor-Native Architecture: Uses multidimensional arrays (tensors) as the core model with Icechunk for cloud-optimized, transactional storage in object stores (AWS, GCP, Azure), unlocking performance and cost savings over table-based or "unstructured" systems.[1][4][5]
• End-to-End Platform: Arraylake enables git-style versioning, sharing, and auditing for teams; Flux provides scalable, standards-compliant APIs for tools like Python, dashboards, and production delivery—eliminating I/O bottlenecks for AI/ML, analytics, and forecasting.[1][4]
• Open-Source Foundation: Built on Xarray, Pangeo, Zarr, and Icechunk (team-led projects), fostering trust, interoperability, and community adoption while supporting major clouds.[1][5]
• Proven Efficiency: Customers report streamlined workflows for terabyte/petabyte-scale data in weather forecasting, environmental monitoring, wildfire risk, and renewables supply/demand—e.g., Sylvera eliminated data bottlenecks.[4][5]

Role in the Broader Tech Landscape

Earthmover rides the surge in physical AI, climate tech, and real-time Earth observation data, where weather/climate datasets explode from static models to dynamic, petabyte-scale assets critical for renewables, insurance, and carbon markets.[1][5] Timing aligns with AI/ML demands for tensor data (rasters/arrays) and cloud-native stacks, as traditional tools like Snowflake fail on non-tabular scientific formats—positioning Earthmover to modernize "deep sciences" (climate, biotech, geospatial) akin to the Modern Data Stack's business impact.[2][5] Market tailwinds include open-source momentum (e.g., Pangeo at Fortune 500s) and investor focus on climate infrastructure, with Earthmover influencing ecosystems by open-sourcing tools that lower barriers for startups and researchers tackling sustainability.[1][2][4]

Quick Take & Future Outlook

Earthmover is primed to scale as the go-to infrastructure for tensor-driven physical AI, expanding tools atop its storage core with $7.2M seed funding to capture growing demand from renewables and climate firms handling real-time forecasts.[1][5] Trends like exabyte-scale Earth data, multimodal AI, and regulated carbon markets will propel adoption, potentially evolving Earthmover into a Snowflake-like powerhouse with broader "deep science" reach (biotech, ML). Its open-source roots and PBC structure ensure ecosystem influence, empowering data to address planetary challenges—just as its founders envisioned from Pangeo origins.[1][2][4]