Sasha Luccioni

Sasha Luccioni is the researcher who made the AI industry’s energy bill a public issue. A machine-learning scientist with a PhD in cognitive computing who trained under Yoshua Bengio at Mila, she leads AI and climate work at Hugging Face and was named to the TIME100 AI list in 2024 for quantifying what generative models actually cost the planet.

Counting AI’s carbon

Luccioni helped create CodeCarbon in 2020, an open-source tool — downloaded more than a million times — that lets developers measure the emissions of running their code. Her research found that generative AI queries use roughly 30 times more energy than a traditional search engine, and that generating a single high-definition AI image can consume as much energy as fully recharging a phone. She notes that Google’s greenhouse-gas emissions rose 48 percent from 2019 and Microsoft’s 29 percent from 2020, driven substantially by AI, despite both companies’ carbon-neutrality pledges.

Her answer is measurement rather than abstinence: the AI Energy Score project rates models the way appliances get efficiency labels. As she put it to AFP, even without accounting for water or rare minerals, “at least we know that for a specific task, we can measure energy efficiency and say that this model has an A+, and that model has a D.” She works inside an AI company — Hugging Face — but her public role is the warning: she argues the industry’s scaling race is environmentally reckless and that companies hide the numbers that would prove it.