Every time you ask an AI chatbot a question, you're consuming roughly 10 times more electricity than a standard Google search, and that energy demand is growing faster than most people realize. A single query to generative AI models like ChatGPT (GPT-4) uses approximately 0.0034 kilowatt-hours of electricity, compared to just 0.0003 kilowatt-hours for a Google search. While these numbers sound tiny in isolation, they add up dramatically when multiplied across billions of daily queries.

The environmental footprint of artificial intelligence depends largely on where the electricity powering data centers comes from. In Canada, where the electricity grid averages roughly 75 grams of carbon dioxide equivalent per kilowatt-hour, each AI query generates approximately 0.3 grams of carbon dioxide equivalent. To put that in perspective, it's equivalent to running a 10-watt LED lightbulb for 15 minutes.

How Much Water Does AI Actually Consume?

Beyond electricity, AI systems have another hidden environmental cost that often goes unnoticed: water consumption. Data centers rely on water-based cooling systems to prevent servers from overheating, and as demand for AI services increases, so does the need for cooling infrastructure. Each AI query takes about 0.26 milliliters of water, which sounds negligible at first, roughly equivalent to five drops. However, when a data center operates in an already dry climate and processes millions of queries every day, the cumulative impact becomes disproportionately higher.

The rapid development of AI technologies also drives demand for specialized hardware, including graphics processing units (GPUs) and servers. Manufacturing and replacing these components contributes to resource depletion and electronic waste, creating environmental costs that extend far beyond the electricity consumed during operation.

What Can Users Do to Reduce AI's Environmental Impact?

The objective isn't to avoid AI entirely, but rather to use it thoughtfully and efficiently to maximize benefits while minimizing environmental impact. Researchers at the University of Manitoba have identified several practical strategies that students, faculty, and staff can implement immediately.

• Think Before You Prompt: Use AI when it provides meaningful value rather than for every task. Avoid using AI for routine queries where a simple web search would suffice, and consolidate multiple requests into a single prompt rather than submitting numerous separate queries.
• Be Specific in Your Requests: Detailed prompts reduce the number of requests required to achieve the desired outcome. Ask for brief, concise summaries to avoid lengthy responses, and use text instead of images when possible, since generating images generally requires more computational resources than generating text.
• Disable AI Features in Your Browser: Turn off AI queries when using browser search by entering '-ai' before your search term to see website results instead of AI summaries. In Microsoft Edge, go to settings, click Copilot and AI, and turn off functions where Copilot uses browser content or takes actions on your behalf.
• Opt Out of Data Training: Google uses your searches to train AI models. To opt out, visit your Google My Activity page, navigate to Search Services History, and select the option to opt out of data training.
• Edit Existing Outputs: Refine or modify a previous response instead of generating a completely new one, and delete unnecessary files and duplicates to reduce overall digital storage demands.

The difference between efficient and inefficient prompting can be substantial. A vague request like "Find me information about dogs" requires the AI to generate a broad response, consuming more energy than a specific prompt like "Find me a short, concise summary of dogs that have short hair, do not shed or are hypoallergenic and are under 30 pounds." Similarly, asking the AI to "rewrite this paragraph" is less efficient than requesting it to "rewrite this paragraph to be more concise with an energetic and friendly tone".

To contextualize the scale of AI's energy consumption, consider that a single Google search query uses approximately 0.0003 kilowatt-hours of electricity. With an estimated 3.5 billion searches conducted each day globally, that adds up to more than 1.05 million kilowatt-hours of electricity daily. By comparison, the average Canadian household consumes roughly 30 kilowatt-hours per day, meaning global daily Google searches consume energy equivalent to roughly 35,000 households.

As AI technologies become more integrated into education, research, and administration across universities and workplaces, the cumulative environmental impact will only grow. By using AI thoughtfully, asking better questions, reducing unnecessary computing demands, and considering environmental impacts in decision-making, users can contribute to a more sustainable digital future while still benefiting from AI's powerful capabilities.