Earthmover continued to sharpen its position as an AI-ready climate and geospatial data infrastructure provider this week, emphasizing both platform capabilities and underlying engineering upgrades. Across several updates, the company highlighted how its cloud-native Marketplace is lowering barriers to working with massive climate, weather, and Earth observation datasets.
Claim 55% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
In a webinar-focused update, Earthmover showcased Planette AI’s hybrid physics-and-AI model for wind forecasting, which reportedly delivers roughly twice the skill of the CFSv2 model for one-month-ahead predictions on a 25 km grid. The session also featured Columbia Climate School’s Bias-Corrected Downscaled Massive Ensemble, a roughly 100 TB, ~1,400-member climate projection dataset extending to 2100.
These resources are being surfaced through Earthmover’s cloud-native data layer, aimed at researchers, energy firms, insurers, and other institutional users seeking scalable access to advanced climate analytics. By enabling both short-term and multi-decadal forecasting use cases on the same infrastructure, the company is positioning itself as critical plumbing for scientific AI and climate risk modeling.
Earthmover also detailed reliability and scale improvements in Icechunk V2, its tensor-data storage technology optimized for cloud object stores. The upgrades target demanding environments with tens of thousands of commits, more than 100,000 arrays, and multi-terabyte, always-on distributed write pipelines where consistency and fault tolerance are essential.
To validate Icechunk V2 at scale, Earthmover is employing open-source tools such as AWS Labs’ Shuttle for exploring asynchronous task orderings, Proptest for randomized repository operations, and Shopify’s Toxiproxy to simulate adverse network conditions. This intensive testing regime is intended to harden network handling, retry logic, and concurrency behavior for enterprise-grade workloads.
The company additionally expanded its AI-ready geospatial offering by adding Taylor Geospatial’s “Fields of the World” global field-boundary dataset to its marketplace. Combined with Sentinel-2 median mosaics and AlphaEarth Foundations embeddings, Earthmover now presents a more complete agricultural AI stack suitable for use cases from smallholder plots to large commercial farms.
Engineering work on the Icechunk array format, designed to integrate with Zarr and Xarray workflows and validated with property-based testing, further underscores Earthmover’s open-source–oriented strategy. On the commercial side, outreach to commodity traders and a wildfire underwriting case study with Kettle demonstrate traction in high-value verticals.
Taken together, the week’s developments highlight Earthmover’s focus on scaling advanced climate and geospatial data access, strengthening its core infrastructure, and deepening ties to data-intensive industries, supporting its longer-term prospects in climate-tech and AI infrastructure markets.