Observed Signal · Jun 18, 2026 · Research Publication · Source: t3n · Impact: 3/5 · Sentiment: Negative
AI Expansion Could Emit 2.8 Gt CO2 by 2031
A study by climate scientist Yassine Charabi, published in Communications Earth & Environment, models the future carbon footprint of scaling artificial intelligence. Using more than 10,000 simulations that draw on IEA energy scenarios, datacenter growth rates and hardware-replacement schedules, the median scenario produces about 2.8 gigatonnes of CO2. In the fastest-case simulations, AI only offsets the emissions from its production by late 2031; until then Charabi describes a period called the "Carbon Valley" where AI systems cause more emissions than they save. The study finds that each year of delay in integrating AI into low-carbon technologies costs roughly 0.45 gigatonnes of CO2. Charabi recommends prioritising embedding AI into environmentally friendly processes to avoid large, hard-to-reverse cumulative emissions that would jeopardise a 1.5 °C climate target.
Quantifies large potential CO2 emissions from rapid AI scale-up and highlights timelines when AI could become net‑carbon‑reducing; relevant to data‑center planning, sustainability commitments and potential regulation affecting AI infrastructure and compute-heavy services.
Track TargetVideo Signals & Market Shifts in Real-Time
Polaris7 autonomous intelligence agents track regulatory filings, primary sources, executive changes, and deal flow 24/7. Create your free Explorer workspace to monitor these entities.
Key Takeaways & Evidence Grounding
- Study by climate scientist Yassine Charabi published in Communications Earth & Environment.
- More than 10,000 model simulations produced a median cumulative CO2 output of 2.8 gigatonnes associated with AI expansion.
- In the fastest-case scenario AI only offsets its production-related impacts by late 2031; before that period is labelled the "Carbon Valley".
- The study estimates each year of delay in integrating AI into clean technologies costs about 0.45 gigatonnes of CO2.
- The model used inputs including global energy forecasts, datacenter growth rates, hardware replacement schedules, daily power consumption estimates and chip production emissions, and referenced IEA scenarios to 2035.
Connected Companies & Entities
1 Entity mappedOntology Mapping & Concepts
Related Market Signals & Shifts
Recent verified developments and strategic activity across this market segment.
What AI Really Costs the Planet
This feature examines the environmental footprint of modern AI, arguing that visible interfaces hide substantial energy, water and materials costs. It compiles recent studies and estimates: data centres already consume a material share of US electricity, training and deploying large models requires large energy and cooling inputs, and manufacturing and rapid turnover of AI-specific GPUs add to carbon emissions and e-waste. The piece highlights uneven geographic impacts (data-centre growth stressing local grids and water supplies), a transparency gap in corporate reporting, nascent regulatory moves in the US and EU, and industry bets on nuclear power as a long-term supply solution. It also discusses psychological effects (eco-anxiety) and behavioral levers like carbon labelling and intentional AI usage as partial mitigations, while noting efficiency improvements in inference that coexist with rapidly rising aggregate demand.
AI's Growing Impact on Earth's Natural Resources
The article analyzes how AI infrastructure and applications consume and transform physical resources — electricity, water, minerals, land, and built infrastructure — through data centres, hardware manufacturing, cooling, and logistics. It cites the IEA estimate that data centres and transmission networks used roughly 460 TWh in 2022 and that consumption could roughly double under high-growth assumptions. The piece explains key operational and embodied metrics (PUE, WUE, kWh per training run, kWh per 1,000 inferences, embodied carbon), highlights mineral and e‑waste risks from rapid hardware turnover, and discusses rebound effects where efficiency gains can increase total demand. It concludes with engineering, procurement, siting, and policy measures to reduce resource pressure, including smaller/specialized models, quantization, scheduling, longer hardware life, water-aware siting, and stricter disclosure and procurement standards.
Google and Amazon Reports Reveal AI's Rising Emissions
Google and Amazon released sustainability reports showing material increases in their carbon footprints as AI usage and associated infrastructure expand. Google’s total emissions rose 25% year-over-year and Amazon’s rose 16%. Much of the increase is driven by Scope 3 emissions tied to capital goods — notably data-center construction, GPUs and semiconductor supply chains — rather than direct energy purchases, which have been moderated by renewable contracts. The reports note growing reliance on fossil-fuel-backed capacity (including investments in natural gas plants) to meet AI power demands. Both companies retain net-zero pledges but face higher costs and tougher decarbonization challenges, including scaling low-carbon steel/cement, expanding renewables, and buying large volumes of carbon removal credits.
Track Real-Time Market Signals & Shifts
Set up custom watchlists to receive automated, evidence-grounded executive digests whenever material signals or shifts occur across your tracked landscape.