PEARC26: ABS-1-4: Addressing the Challenges Posed by Anaconda’s New Licensing Model

When Anaconda Inc. introduced its revised Terms of Service (ToS) in Spring 2024, the changes created immediate uncertainty for academic research computing. For years, Anaconda’s Python distribution and package‑management platform had been freely available for academic use—both on local systems and across HPC clusters—serving as the foundation for scientific workflows in many departments. Institutions that rely on shared local systems and HPC clusters faced concerns about compliance scope, budget exposure, and potential disruption to established collaborative workflows. This work reflects that context while focusing on practical and current paths forward. The conda package manager is free and open-source — it is not going away. The licensing concerns are specifically about packages hosted in Anaconda Inc.’s “defaults” channel and their distribution platform. **For most users, the simplest path forward is to use conda-forge as your package channel (e.g., by installing Miniforge instead of the Anaconda distribution). Conda-forge is a large, community-maintained collection of packages that does not fall under Anaconda’s Terms of Service. You can continue using conda (or mamba) exactly as before — just sourcing packages from conda-forge instead of “defaults.” That said, Anaconda Inc. does provide value beyond just hosting packages. Their curated “defaults” channel undergoes additional vetting, which may matter for institutions with stringent security or reproducibility requirements. Organizations in regulated or high-security environments may find Anaconda’s commercial offerings — including verified package provenance and enterprise support — worth evaluating. For others, conda-forge covers most scientific computing needs at no cost and with no licensing concerns. Anaconda Inc.’s revised Terms of Service, effective July 2025, introduce new compliance obligations for academic institutions. While Anaconda remains free for accredited educational entities under their Academic Policy, institutions must now proactively secure an Eligible Academic Institution (EAI) agreement — renewed annually — and are held responsible for ensuring all users comply with the terms. Key risk areas include restrictions on embedding Anaconda packages in containers, mirroring repositories, and providing access to third parties (e.g., external collaborators), all of which may require a paid license even under the academic exemption. The terms can be modified with as little as 90 days’ notice, and it falls to the institution to monitor for changes. Given this evolving landscape, the working group recommends that institutions evaluate open-source alternatives (such as Miniforge with conda-forge) to reduce compliance exposure, while engaging with peer institutions through CaRCC’s Anaconda Transition Working Group to advocate for clearer, stable academic terms.

Authors: Manasvita Joshi, Geoffrey Lentner, Aldo Romero, Anna Alber, Bryan Raney, Ian Kaufman, Joseph Creech, Lev Gorenstein and Raffaella D D’Auria

Room: Room 101A & 101B