Artificial intelligence models have a high carbon and environmental footprint
the verdict
SUPPORTED
the evidence backs this
confidence 86/100
Current research consistently indicates that artificial intelligence models, particularly large language models and generative AI systems, carry a substantial carbon footprint, high energy demands, and significant environmental impacts across their hardware and operational lifecycles.
Evidence for · 9
Reconciling the contrasting narratives on the environmental impact of large language models
2024 · cited by 57
Acknowledges that LLMs have substantial environmental impacts through energy consumption, carbon emissions, and water usage.
See more details
More for · 8
Towards Green AI in Fine-tuning Large Language Models via Adaptive Backpropagation
2023 · cited by 24
Highlights that intensively performed LLM fine-tuning worldwide results in significantly high energy consumption and carbon footprint.
Large Language Models (LLMs): Deployment, Tokenomics and Sustainability
2024 · cited by 14
Discusses sustainability challenges and environmental carbon footprint impacts associated with state-of-the-art LLMs.
Sustainable AI and Green Computing: Reducing the Environmental Impact of Large-Scale Models with Energy-Efficient Techniques
2025 · cited by 8
Explores concerns over the substantial energy consumption and associated carbon emissions of large-scale AI models.
An Introduction to Life-Cycle Emissions of Artificial Intelligence Hardware
2025 · cited by 5
Provides a comprehensive life-cycle assessment of AI hardware, showing that manufacturing and operating specialized accelerators carry a heavy environmental cost.
AI’s Thirst, AI’s Heat, AI’s Waste: Exposing the Hidden Environmental Impact of Every Artificial Intelligence Interaction
2025 · cited by 4
Demonstrates that generative AI has a severe physical footprint, driven by intensive operational inference, water consumption for cooling, and e-waste.
Sustainable generative AI and quantum computing: review assessment on the environmental impact of generative AI and quantum technologies
2026 · cited by 1
Confirms that the ecological footprint of generative AI is driven by immense energy demands for large-scale model training and inference.
Emissions and Performance Trade-off Between Small and Large Language Models
2025 · cited by 0
Notes that the advent of LLMs has raised serious concerns about their enormous carbon footprint starting from energy-intensive training and repeated inference.
Artificial Intelligence’s Hidden Footprint in Environmental Engineering: A Life-Cycle Risk Assessment and the AI-ERAF Governance Framework
2025 · cited by 0
Examines how training large deep learning models pumps out hundreds of metric tons of CO2 and strains resources through electronic waste and mining.