Environmental Impacts of AI

The Environmental Cost of AI

Energy Consumption

• Training large AI models consumes extraordinary energy
• A single ChatGPT query uses 5-10x more electricity than a Google search
• By 2026, data centers could use up to 14% of Canada’s power

Water Usage

• Data centers need 9 liters of water per kWh for cooling
• GPT-3 training required approximately 700,000 liters of water
• Much of this water is lost through evaporation

What Can Be Done?

Researchers and companies are working on “Making AI Less Thirsty” through:

• More efficient model architectures
• Better data center cooling technologies
• Renewable energy for training and inference
• Smaller, more efficient models for common tasks

For Library Professionals

Be mindful of unnecessary AI use. Consider whether you really need AI for a task, or if simpler tools would suffice.

Video content coming soon – placeholder for video embed