Earthmover PBC

Earthmover PBC is a United States-based public benefit corporation, headquartered in New York, that operates a cloud data platform for scientific data. The company describes itself as "the science data cloud" and states its mission as empowering people to use scientific data to solve humanity's greatest challenges. Its platform is designed to make the world's scientific data accessible for both humans and AI, and it is trusted by leading teams and data providers according to its own homepage. The company was founded by Ryan Abernathey and Joe Hamman, both of whom began their careers as climate scientists. Abernathey, a professor at Columbia University, worked with petabytes of climate data and encountered the limitations of existing tools and infrastructure. Hamman faced similar challenges at the National Center for Atmospheric Research (NCAR) and CarbonPlan. Their shared commitment to open-source development brought them together, and they founded Earthmover to build foundational data infrastructure that could accelerate scientific discovery. Earthmover's platform is organized around three core capabilities: storage, governance, and compute. For storage, the company developed Icechunk, an open-source (Apache 2.0) technology that represents scientific data as infinitely extensible multidimensional arrays — chunked on any axis, compressed, and versioned. Icechunk provides zero-copy ingest that turns NetCDF, HDF5, GRIB, and TIFF archives into structured data cubes, with ACID transactions for operational workloads. Earthmover reports that when NASA implemented Icechunk, they saw a 100x speedup. For governance, the company offers Arraylake, a metadata and governance layer providing catalog, search, ownership, permissions, and complete version history over existing data. Arraylake supports bring-your-own-bucket or managed storage, and works with every public cloud as well as S3-compatible on-prem object storage. For compute, Earthmover runs workloads next to the data in the same cloud region, on hardware that scales out to match the size of the problem, making repositories queryable over standard protocols. Earthmover positions itself as a trusted partner to industry, government, and academia, and aims to contribute broadly to a multi-stakeholder open-source ecosystem. The company also states an intention to cultivate a healthy and inclusive work culture. Its vision is a world in which scientific data can be explored, visualized, analyzed, shared, and built upon effortlessly, providing a foundation for new scientific knowledge and data-driven products.

Earthmover was founded by Ryan Abernathey and Joe Hamman, two climate scientists who independently encountered the difficulties of working with massive scientific datasets. Abernathey, as a professor at Columbia University, analyzed petabytes of climate data and pushed against the limitations of existing tools and infrastructure. Hamman encountered similar challenges during his work at the National Center for Atmospheric Research (NCAR) and CarbonPlan. Both became active contributors to the open-source community, building projects to help scientists work more effectively with large-scale data. Their paths crossed through this shared open-source work, and they discovered a common vision: to build foundational data infrastructure that could accelerate scientific discovery and help address humanity's most pressing challenges. That vision led to the creation of Earthmover PBC, a public benefit corporation headquartered in New York, United States. The company built a cloud data platform organized around three pillars: storage, governance, and compute. Its storage technology, Icechunk, was developed to solve the problem that most scientific data sits in opaque binary formats spread over millions of individual files with no index and no map, making cloud queries require downloading each file in full. Icechunk represents data as infinitely extensible multidimensional arrays, chunked on any axis, compressed, and versioned, with zero-copy ingest from NetCDF, HDF5, GRIB, and TIFF archives and ACID transactions for operational workloads. Earthmover reports that NASA's implementation of Icechunk produced a 100x speedup, and the technology is released as 100% open source under the Apache 2.0 license. The governance layer, Arraylake, addresses the problem that scientific data is often essentially invisible within organizations, living in random buckets with no discoverability or governance. Arraylake provides catalog, search, ownership, permissions, and complete version history over data that organizations already have, supporting bring-your-own-bucket or managed storage and working with every public cloud and S3-compatible on-prem object storage. The compute layer runs workloads next to the data in the same cloud region, on hardware that scales out to match the size of the problem, so that any repository becomes queryable over standard protocols. This design reflects the company's view that at petabyte scale, moving data is the slowest and most expensive operation, and that different consumers expect data in different shapes and over different protocols. Earthmover states that its platform is trusted by leading teams and data providers and that it is accelerating scientific progress by making the world's scientific data accessible for humans and AI. The company positions itself as a key trusted partner to industry, government, and academia while contributing broadly to a multi-stakeholder open-source ecosystem, and it aims to cultivate a healthy and inclusive work culture.

• Ryan Abernathey and Joe Hamman meet through shared open-source development work.

• Earthmover PBC is founded as a public benefit corporation headquartered in New York, United States.

• Icechunk is developed and released as 100% open source under the Apache 2.0 license.

• NASA implements Icechunk and sees a 100x speedup.

The science data cloud — a platform making scientific data accessible for humans and AI, trusted by leading teams and data providers.

• Icechunk

  Open-source (Apache 2.0) storage technology that represents scientific data as infinitely extensible multidimensional arrays — chunked on any axis, compressed, and versioned. Provides zero-copy ingest from NetCDF, HDF5, GRIB, and TIFF archives into structured data cubes, with ACID transactions for operational workloads.

• Arraylake

  Metadata and governance layer of the platform, providing catalog, search, ownership, permissions, and complete version history over existing data. Supports bring-your-own-bucket or managed storage, and works with every public cloud and S3-compatible on-prem object storage.

• Cloud compute services

  Compute runs workloads next to the data in the same cloud region, on hardware that scales out to match the size of the problem, making any repository queryable over standard protocols.

1. 1Earthmoverearthmover.io
2. 2About | Earthmoverearthmover.io
3. 3Earthmover PBC (United States) - Wikidatawikidata.org