top of page

Unveiling the Hidden Cost: How AI's Water Footprint Impacts Our Planet


What if the simple act of typing a question into an AI chatbot could be silently draining our planet's precious water resources?

As we celebrate World Environment Day, it’s time to uncover the hidden environmental costs of our technological conveniences. Artificial intelligence (AI) has transformed industries and personal interactions, but this progress comes with significant resource demands. Leading companies like Microsoft and OpenAI, at the forefront of AI innovation, require enormous amounts of water to cool their data centers. At Yzerly, a pioneer in communication training, we are committed to sustainability. We believe that by refining our AI prompt writing techniques, we can reduce water consumption, making our technological advancements more environmentally friendly.

Ai's water footprint

Market Insights

The environmental impact of AI, particularly in terms of water consumption, is becoming increasingly evident.

Recent studies have highlighted the significant water usage required to run AI models like ChatGPT. For instance, training the GPT-3 model at Microsoft’s data centers consumed approximately 700,000 liters (around 185,000 gallons) of water. This staggering amount is equivalent to producing about 370 BMW cars or 320 Tesla electric vehicles​ (The Independent)​​(ar5iv)​.

The operational water usage of AI models is equally noteworthy. It's estimated that a conversation involving 20 to 50 questions with ChatGPT requires about 500 milliliters (or a standard 16.9-ounce bottle) of water. While this may seem minimal per interaction, the cumulative effect, given ChatGPT's vast user base, results in immense water consumption. The water is primarily used for cooling servers and maintaining necessary humidity levels within data centers, as saltwater is not a viable option due to corrosion and other complications​ (The Markup )​​ (Yahoo)​.

In response to these findings, companies like Microsoft and OpenAI are under increasing pressure to address their environmental impact. Both companies have acknowledged the issue and are exploring ways to enhance their data centers' efficiency and reduce water consumption. Microsoft, for example, has committed to becoming "water positive" by 2030, aiming to replenish more water than it consumes​ (The Independent)​​ (Yahoo)​.

Furthermore, the global demand for data centers to support AI technologies is projected to significantly impact water resources. By 2027, data centers could require water withdrawals of 4.2 to 6.6 billion cubic meters​ (The Markup )​. This has led to a push for innovative cooling solutions. Companies like VIRTUS Data Centres and Submer are exploring new technologies to cut water usage by up to 55%, such as immersion cooling where servers are submerged in a thermally conductive but electrically insulating liquid​ (ar5iv)​.

Overall, while AI technologies like ChatGPT offer significant advancements, their environmental footprint, particularly regarding water consumption, poses a critical challenge. This necessitates innovative solutions and greater transparency from the companies involved to mitigate these impacts effectively​ (ar5iv)​​ (The Markup )​​ (Yahoo)​.

The Challenge

The primary challenge in addressing the environmental impact of AI is raising awareness about the significant water consumption associated with AI model operations, particularly the length and frequency of user interactions with systems like ChatGPT.

AI models like ChatGPT require substantial water resources to cool data centers and maintain optimal operating conditions. Every conversation with ChatGPT, involving 20 to 50 questions, uses approximately 500 milliliters (16.9 ounces) of water​ (The Markup )​​ (Yahoo)​. While this might seem minor on an individual level, the cumulative effect across millions of users results in immense water consumption.

The challenge is to make users and companies aware of this hidden water footprint. Many users are unaware that their frequent and prolonged interactions with AI models contribute to significant water usage. This lack of awareness can impede efforts to reduce the environmental impact of AI technologies.

Educating users on the importance of efficient AI usage is crucial. By writing better prompts, users can get quicker and more accurate responses, reducing the need for prolonged interactions. This can lead to a substantial reduction in the water required for cooling servers during these interactions. Additionally, promoting transparency and efficiency in AI operations can help users make more informed decisions about their AI usage.

In summary, the key challenge lies in making people aware of the environmental costs associated with their AI interactions. By understanding the water footprint of AI models and adopting efficient usage practices, users can play a vital role in mitigating the environmental impact of these technologies.

The Solution

Addressing the environmental impact of AI requires not only technological innovations but also improvements in how we interact with these systems. At the heart of this is better communication skills, which are directly correlated with effective prompt engineering.

  • Better Communication, Better Prompts Effective communication is essential in all areas of interaction, and this holds true for human-AI interactions as well. By learning to craft clear, concise, and precise prompts, users can significantly reduce the computational resources needed to generate accurate and relevant responses. This not only enhances the efficiency of AI systems but also helps in reducing their environmental footprint.

  • The Role of Prompt Engineering Prompt engineering involves creating well-structured and effective inputs for AI models to process. Good prompt engineering can lead to quicker, more accurate responses, thereby reducing the time and energy AI systems need to operate. This, in turn, reduces the water consumption required for cooling data centers. When users provide vague or complex prompts, AI systems need to perform more computations to understand and respond correctly, increasing resource consumption.

  • Yzerly's Initiative At Yzerly, we conduct in-house training to our employees to master the art of effective communication, including prompt engineering for AI applications. Our workshops emphasize the principles of clarity, brevity, and relevance—key components of successful prompt engineering. By honing these skills, participants learn to create efficient AI interactions, which can substantially reduce the length and complexity of AI-driven conversations.

  • Practical Training Our prompt engineering workshops provide hands-on training on how to formulate questions and commands that AI models can process efficiently. Participants learn to:

    • Be Specific: Narrowing down the scope of questions to eliminate ambiguity.

    • Use Clear Language: Avoiding jargon or overly complex language that can confuse AI models.

    • Structure Prompts Properly: Organising information logically to aid the AI in understanding and processing the request swiftly.

Through these practices, users can achieve faster and more relevant responses from AI models, minimising the environmental impact associated with their use.

Sustainable AI Practices

In addition to training, we advocate for broader sustainable practices within the AI industry. This includes encouraging companies to invest in innovative cooling solutions and to be transparent about their environmental footprints. Technologies such as immersion cooling, where servers are submerged in a thermally conductive but electrically insulating liquid, are promising advancements that can reduce water usage by up to 55%​ (ar5iv)​.


At Yzerly, we are not opposing the use of ChatGPT or other AI tools. On the contrary, we recognize their immense benefits and transformative potential across various sectors. However, it's crucial to acknowledge and address the environmental footprint of these technologies. By becoming better at communicating with AI, specifically through crafting smaller, more effective prompts, we can significantly reduce water usage associated with AI operations.

Improving our prompt engineering skills means shorter, more precise interactions with AI models. This not only enhances the efficiency of these tools but also minimizes the computational resources required, leading to less water consumption for cooling data centers. As we continue to leverage AI for innovation and productivity, let's also commit to using these technologies responsibly and sustainably. Together, we can make a positive impact on our environment while enjoying the benefits of AI.

9 views0 comments


bottom of page