Calculadora de Água de IA (ChatGPT etc.)
- Created by
- Renato Passos, Eng. de Software
- Reviewed by
- Renato Passos, Eng. de Software
Last updated: Apr 18, 2026
Formula
Água ≈ 50 mL por consulta (baseado em ~500 mL / 10 consultas GPT-4)
About this calculator
This calculator estimates the water consumption associated with the use of generative artificial intelligence, such as ChatGPT, GPT-4, and similar models. Each query to these models consumes water to cool the servers in data centers. The estimate is based on research indicating that a conversation with 10 GPT-4 queries can consume about 500 mL of water. Thus, each query equals approximately 50 mL.
The calculation is simple: enter the number of queries you make per day or per month, and the tool multiplies that value by 50 mL. The result is presented in liters, milliliters, and a comparison with water bottles. For example, 100 daily queries result in 5 liters of water consumed indirectly, equivalent to about 10 500 mL bottles.
This tool is useful for raising awareness about the environmental impact of artificial intelligence. While AI brings benefits, its intensive use in data centers consumes significant water resources. By estimating your consumption, you can reflect on more sustainable practices, such as optimizing queries or choosing times of lower energy demand. The calculation is an approximation, as actual consumption varies depending on the data center location and cooling system efficiency.
Caveats: the 50 mL per query value is an average; newer models or more complex tasks may consume more. Additionally, the water used is not necessarily lost, as it can be treated and reused, but it still represents resource use. Use this calculator as an educational tool, not an exact measurement. For a more precise impact, consult sustainability reports from AI providers.
Frequently asked questions
Why does AI consume water?
Servers running AI models generate a lot of heat and need cooling. Many data centers use water for this cooling, evaporating or discharging it after use.
Is this consumption really significant?
Yes, especially with the growth of AI usage. Studies show that global water consumption by data centers can be comparable to that of a medium-sized country.
How can I reduce my impact?
Avoid unnecessary queries, reuse responses, and prefer times of lower energy demand. Also support providers using renewable energy and efficient cooling systems.
Is the calculation accurate for all AI models?
No. The 50 mL per query value is an average based on GPT-4. Smaller models may consume less, and larger models or those with longer responses may consume more.
Is the consumed water lost forever?
Not always. Some water evaporates during cooling, but another part can be treated and reused. Still, the process consumes water resources and energy.