Climate14 November 2026What share of global emissions comes from AI and digital technology?
Digital technology and AI as a share of global greenhouse gas emissions, with what each figure counts
The Borregaard plant in Sarpsborg, Norway. Photo Vegard Gran, CC BY-SA 4.0, via Wikimedia Commons.
Digital technology accounts for 1.4 to 3.9% of global greenhouse gas emissions according to studies, which count screens, networks and data centres. AI is a small part of it: 99 MtCO2e in 2025 according to GreenIT, about 0.2% of global emissions. Its share could reach about 1% in 2030 in the highest estimate, the one that assumes servers in continuous use.
Contents
In 2024, humanity emitted 57.7 billion tonnes of greenhouse gases (GtCO2e), a record, according to the United Nations Environment Programme (UNEP), forests and soils included. The AI shares that follow are calculated against this total (our calculation); the digital technology shares are published by each study, against global emissions in the year of its data.
Share of global greenhouse gas emissions, in %: digital technology according to 3 studies, AI according to GreenIT set against UNEP's 57.7 GtCO2e for 2024 (our calculation).
The 2 shares do not measure the same thing. The digital technology share counts computers, phones, networks and data centres, from manufacture to use. The AI share counts only AI servers, their storage, their network and the centres that house them, excluding users' devices. AI is therefore included in digital technology, but the studies of digital technology predate its rise and barely count it.
Digital technology: 1.4 to 3.9% depending on what is counted
4 estimates are the benchmark. They cover nearby years, from 2019 to 2023:
- 3.5 % in 2019 for The Shift Project, or 1.84 GtCO2e, televisions included;
- 3.4 % in 2023 for GreenIT, or 1.8 GtCO2e in life cycle assessment;
- 2.1 to 3.9% in 2020 for Freitag and co-authors, who correct earlier studies for the supply chain links they overlooked;
- 1.4 % in 2020 for Malmodin and co-authors, researchers at Ericsson and Telia, or 764 MtCO2e, excluding televisions and media, counted separately at about 350 MtCO2e.
The gap comes first from scope. With televisions, the estimates range from 2.1 to 3.9%; without them, Malmodin and his co-authors arrive at 1.4%. The next chart places the 4 estimates side by side, with the year of their data.
ClimateThe digital sector in global emissions
Part du numérique dans les émissions mondiales de gaz à effet de serre, en % · Données 2019 à 2023
Source : Shift Project ; GreenIT ; Freitag et al. ; Malmodin et al. · projetcelsius.com/en/ai · CC BY 4.0
AI: about 0.2% in 2025
The only global life-cycle estimate of AI emissions that we know of is the one by GreenIT, published in September 2026: 99 MtCO2e in 2025, server manufacturing included. Set against global emissions of 57.7 GtCO2e, this makes 0.17 % (our calculation), as much as the emissions of Belgium. The study counts:
- 1.9 million AI servers in 2025, or 13 million GPUs;
- their storage, their network and the data centres that house them;
- but neither users' computers and phones nor access networks.

What the International Energy Agency counts
L'AIE counts the electricity of all data centres, AI included: about 180 million tonnes of CO2 in 2024, or 0.5% of CO2 emitted by combustion worldwide, excluding server manufacturing and backup generators. The 2 figures cannot be added together or compared term by term: one counts the life cycle of AI servers, the other the electricity of all centres. The next chart sets GreenIT's estimate against countries and aviation.
ClimateAI emissions, compared
Annual emissions, in millions of tonnes of CO2e · 2024 to 2030
Source: GreenIT, 2026; Eurostat; Citepa; ATAG · projetcelsius.com/en/ai · CC BY 4.0
In 2030: about 1% in the highest estimate
GreenIT projects 636 MtCO2e for AI in 2030, 6 times more than in 2025, with 6.2 million servers. At constant global emissions, the AI share would rise from 0.17 to about 1.1 % (our calculation). This projection is the highest at present, for 2 reasons:
- it assumes servers used at 69% of their capacity, 24 hours out of 24, all year round;
- it keeps the 2025 electricity mix, without the gains from the renewables and nuclear power being added.
The IEA, for its part, sees emissions from data centre electricity rising from 180 MtCO2 in 2024 to about 320 in 2030 and 300 in 2035 in its base case, while staying below 1.5% of energy sector emissions. Our analysis of the gap between estimates sets out these methodological choices, and the one on the climate crisis covers indirect effects, which are in none of these shares.
Our reading
For us at Projet Celsius, AI weighs little in global emissions today, and data centres are, according to the IEA, among the fastest-growing sources of emissions. Its share remains small in 2025, and it could be multiplied by 6 by 2030 in the highest estimate. Our analysis of the most serious risk puts it at the scale of electricity grids, and the comparison with aviation sets it against a well-known sector. When you see a percentage of global emissions, ask 3 questions:
- Does it count all digital technology or AI alone?
- Does it count the manufacture of devices and servers, or only electricity?
- What year is the data from, and what total is it set against?
The 0.17% and 1.1% figures are our calculations: AI emissions according to GreenIT set against UNEP's 57.7 GtCO2e for 2024. GreenIT does not publish them. The IEA's 180 MtCO2 is CO2 only, for the electricity of all data centres, AI included.
Sources
8 documents
- UNEP, Emissions Gap Report 20252025entry 48
- GreenIT, Impacts environnementaux et sanitaires de l'intelligence artificielle dans le monde en 2025 et 20302026entry 6
- GreenIT, Impacts environnementaux du numérique dans le monde2025entry 7
- The Shift Project, Impact environnemental du numérique : tendances à 5 ans et gouvernance de la 5G2021entry 5
- Freitag et al., The real climate and transformative impact of ICT, Patterns2021entry 40
- Malmodin et al., ICT sector electricity consumption and greenhouse gas emissions - 2020 outcome, Telecommunications Policy2024entry 41
- IEA, Energy and AI2025entry 1
- Eurostat, greenhouse gas emissions (env_air_gge)2026entry 50
Every figure was read in its source; the hub's methodology is described in Methodology and sources.
AI Impact, Projet Celsius, "What share of global emissions comes from AI and digital technology?", 14 November 2026. https://projetcelsius.com/en/ai/analyses/what-share-of-global-emissions-comes-from-ai/















Hardware






