AI Infrascience

The study of the infrastructure that supports AI research and deployment: datasets, computing clusters, software frameworks, model zoos, annotation pipelines, and the energy grids that power them. AI infrascience examines how infrastructure choices shape what AI can do—who gets to train large models, whose data is included, how carbon footprints are measured, and how open source vs. proprietary tools affect innovation. It reveals that AI is not just algorithms; it’s a material system dependent on rare earth minerals, cloud contracts, and precarious labeling labor. Understanding AI requires understanding its infrascience.

Example: “Her AI infrascience research traced how the shift to cloud computing centralized AI power in a few tech giants, making it nearly impossible for academics to compete on large language models.”

by Dumu The Void April 11, 2026