September 2025
Artificial intelligence (AI) can accelerate data analysis, improve decision-making, and transform how nonprofits and public health organizations work. Like any technology, it comes with an environmental footprint. Understanding that footprint and making informed, practical choices helps organizations use AI responsibly and align with sustainability goals.
1. Individual Use Has a Small but Real Impact: Most individual AI use is light on resources. A single ChatGPT query consumes about 0.3 watt-hours of electricity and less than one milliliter of water, roughly the same as running your shower for one second (Masley, 2024).
For individual users or small teams, the carbon and water footprint per interaction is minimal. However, scale changes the story. When organizations embed AI across daily operations and use high-powered models for thousands of queries each day, the cumulative impact becomes more meaningful. It is a reminder that while personal use may be small, organizational use patterns deserve reflection and planning.
2. Match the Tool to the Task: Not every job needs the most powerful model. Many everyday AI tasks such as summarizing a document, rephrasing content, or drafting bullet points can be performed by smaller, more efficient models (CNN, 2025). Tools like Claude 3 Haiku, Gemini Flash, or GPT-3.5 can reduce emissions by 90–99% compared to GPT-4 (MERLTech, 2024).
Think of model choice like transportation: you do not need a semi-truck for a gallon of milk. Lighter models run faster, are often cheaper, and dramatically reduce environmental load. Making intentional model selections is one of the simplest and most effective ways to reduce AI’s environmental footprint without sacrificing productivity.
3. Weigh the Gains Against the Costs: AI has the potential to deliver enormous global benefits. The International Monetary Fund (IMF) estimates that AI could increase global GDP by 0.5 percentage points annually between 2025 and 2030, a major economic lift (Reuters, 2025 & Axios, 2025). However, this growth carries an environmental cost of roughly 1.2 gigatons of CO₂ over five years if the world continues to rely on today’s fossil-fueled energy mix.
The balance depends on how systems are powered and deployed. When powered by renewable energy and applied thoughtfully, AI can create a net environmental benefit. But intentionality matters: the same system used carelessly can deepen energy inequities and emissions.
4. Transparency and Accountability Are Key: Understanding AI’s environmental impact requires reliable and comparable data, but few companies are sharing it. Mistral AI (2025) recently set a new standard by publishing a full lifecycle assessment of its “Mistral Large 2” model, detailing both training and inference impacts:
- Training: 20.4 kilotons of CO₂ equivalent and 281,000 cubic meters of water
- Inference (per 400-token response): 1.14 grams of CO₂ and 45 milliliters of water
This level of disclosure is rare but essential. It gives organizations a benchmark and users the power to make informed choices. If one company can do it, others can and should. Without transparent data, it is nearly impossible to set sustainability standards or track progress.
5. Moving Toward Responsible AI Use: As organizations integrate AI into evaluation, research, and operations, sustainability should be part of the conversation. Practical steps include:
- Right-size your model to the complexity of the task.
- Batch and reuse outputs when possible to reduce redundant processing.
- Choose providers committed to renewable energy or carbon-neutral operations.
- Include sustainability metrics such as energy and water use in vendor selection or AI governance policies.
- Advocate for transparency and shared reporting standards so the field can make better, evidence-based decisions.
Used thoughtfully, AI does not have to conflict with environmental responsibility. It can be part of a more sustainable and efficient future if we design and deploy it that way.
- Masley, Andy. “Individual AI Use Is Not Bad for the Environment.” Substack, 2024. https://andymasley.substack.com/p/individual-ai-use-is-not-bad-for
- MERL Tech. “The Hidden Cost of Our AI Habits.” MERLTech.org, 2024. https://merltech.org/the-hidden-cost-of-our-ai-habits/
- CNN. “Your AI Prompts Are Adding to Carbon Emissions. Here’s What to Know.” CNN Climate, June 22, 2025. https://www.cnn.com/2025/06/22/climate/ai-prompt-carbon-emissions-environment-wellness
- Reuters. “AI Economic Gains Likely Outweigh Emissions Cost, Says IMF.” Reuters, April 22, 2025. https://www.reuters.com/sustainability/climate-energy/ai-economic-gains-likely-outweigh-emissions-cost-says-imf-2025-04-22
- Axios. “IMF: AI Climate Toll Is Big—But the Payoff May Be Bigger.” Axios Future, April 22, 2025. https://www.axios.com/2025/04/22/ai-climate-toll-imf-study
- Mistral AI. “Our Contribution to a Global Environmental Standard for AI.” Mistral Newsroom, April 2025. https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai