Analysis ·
Estimating the footprint of AI use from partial data: a key issue for AI governance
Original title, in French: “Estimer l’empreinte des usages IA à partir de données partielles, un sujet crucial pour sa gouvernance”
- Publication
- Alliancy (France)
- Author
- Antoine Roba
- Published on
- 29 September 2026 · read in the page structured data
Summary
According to the article, hyperscalers' AI investment rises from $395 billion to $530 billion between 2025 and 2026. Generative AI is entering business use through four consumption models (SaaS, on-premise, cloud IaaS/PaaS, office suite add-ons), each with environmental impacts that are more or less measurable. Estimating the carbon footprint relies on the model's life cycle, with the inference phase being the most energy-intensive. Despite the uncertainty of the available data, particularly for SaaS, these partial estimates allow organisations to prioritise actions, raise awareness among teams and discuss environmental commitments with their cloud providers, in a context of tighter European regulation on reporting.
Summary based on the full text of the article.
Read the article on Alliancy →Figures cited
-
530 $bn
Increase in hyperscalers' AI investment (CapEx)
2026
« Les hyperscalers augmentent leurs CapEx de 395 Md$ en 2025 à 530 Md$ en 2026. »
-
4 %
Share of global greenhouse gas emissions attributable to digital technology
global
« Le numérique représente environ 4% des émissions de gaz à effet de serre (GES) mondiales »
-
395 $bn
Increase in hyperscalers' AI investment (CapEx)
2025
« Les hyperscalers augmentent leurs CapEx de 395 Md$ en 2025 à 530 Md$ en 2026. »
Classification
- Country
- Worldwide scope
- Company size
- All sizes
- Industries
- All industries, or industry not specified · Software vendors, IT services companies, telecoms
- Companies
- Carbone 4 (cited), Mistral AI (cited)
- Keywords
- carbon footprint, AI governance, digital sobriety, energy consumption, environmental reporting, generative artificial intelligence
Entry written by Marsanne Studio from the cited article, with the help of an AI model (claude-sonnet-5). The rights to the article and its illustrations belong to Alliancy and its author. Entry translated from French by Claude Sonnet 5.5.