Expanding “Green AI” Beyond Model Efficiency
Year: 2026
Type: article
Abstract: This commentary argues that the artificial intelligence (AI) boom is not immaterial but relies on energy- and resource-intensive infrastructures. While “Green AI” scholarship has advanced model-level efficiency metrics and reporting, it has largely overlooked the material circuits that enable these models. We explain how efficiency-first framings are reinforced by an ideological blend of cybertarianism and techno-nationalism, which together reframe environmental and labor externalities as acceptable costs of competition. We propose three priorities for a more adequate agenda: (1) adopt the “trash metaphor” to highlight the afterlives and externalities of AI infrastructures; (2) implement structural solutions that oversee and regulate the entire material circuits of AI; and (3) foster green citizenship through public information rights, participatory siting, and ongoing civic oversight to advocate for and sustain the necessary structural changes. Overall, we call for aligning AI's claimed climate benefits with demonstrable compliance to ecological budgets rather than aspirational efficiency narratives.
Source: Emerging Media
Authors: Sibo Chen, Deqiang Ji
Language: English
Topic: Green IT and Sustainability
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7160144917