“AI Is Thirsty & Hungry”: the Hidden Environmental Costs of Artificial Intelligence
- The White Hatter
- Jun 14
- 4 min read
At The White Hatter, we talk a lot about the digital risks and opportunities associated with AI, from online privacy to student learning. However, there's another growing concern that’s often left out of the conversation, AI’s environmental footprint and more specifically how much water and power it consumes to function.
As the popularity of AI tools like ChatGPT and generative models soars, so too does their demand on resources. This article explores the lesser-known side of artificial intelligence, the environmental costs, and why it matters for families, educators, and teens to start asking smarter questions, and not just about what AI can do, but what it takes to do it.
Every time you type a question into ChatGPT, you’re triggering a complex web of computer computations. These calculations generate heat, a lot of it, which must be cooled to prevent damage to the servers, and the most common solution is water.
A single 10–50 prompt session with ChatGPT uses about as much water as a standard 16 oz bottle. (1)
Training just one large AI model (like the ones that power ChatGPT) in Microsoft's U.S. data centres evaporated over 700,000 gallons of water. (2)
In 2022, Microsoft used 1.7 billion gallons of water (the equivalent of 2,500 Olympic-sized pools), while Google used 5.6 billion gallons, before ChatGPT became a household name. (3)
This isn't just happening in the U.S. Canada is seeing a rise in “hyperscale data centres,” often built near freshwater sources such those being built in Kamloops BC and Merit BC where Bell is building two AI data centres. (4) These large AI data centres can consume up to 550,000 gallons of water per day, equivalent to the daily water usage of thousands of people. Along with municipalities, homeowners, businesses, manufactures and farmers, AI data centres are now competing for this finite resource.
As an example, the CRD that manages the drinking water in Greater Victoria where we live, estimates that the average daily water use per person is about 58 US gallons per day. (5) If we use the 550,000 stat that was mentioned about the water that these large AI data services need, that equate to the amount of water that 9,483 individuals would consume in one day.
However, AI’s thirst doesn’t stop at water. It also consumes immense amounts of electricity. A single large data center can consume as much power as 50,000 homes per day (6)
With the global push toward electric vehicles, increased air conditioning use due to climate change, and an aging electrical grid, the strain is only getting worse. (7)
According to the research, AI data center power demand is expected to increase 15–20% every year through 2030. That’s a massive growth curve with real-world implications. (8)
In Omaha Nebraska, Google and Meta are each building new AI-focused data centres. A nearby coal plant that was scheduled for shutdown is now being kept online to meet the energy demands, a clear example of how AI’s appetite can delay our transition away from fossil fuels which have real environmental concerns. (9)
Looking back at Kamloops and Merritt, both regions are already grappling with water and power shortages. As demand continues to rise from local communities, agriculture, municipalities, and industry, now joined by the addition of two AI data centres, it’s unclear whether these new facilities will place even greater pressure on already strained critical infrastructure, time will tell.(10)
We believe that the rise of AI data centres will presents a growing tension between economic opportunity and environmental sustainability, what some are calling a new kind of jobs vs. environment tug-of-war. On one hand, these hyperscale facilities bring the promise of jobs, tech investment, and regional economic growth. Communities struggling with employment shortages or the decline of traditional industries may see AI infrastructure as a much-needed lifeline. Governments and tech companies alike are promoting these projects as engines of innovation, often highlighting the direct and indirect employment they create, from construction and maintenance to cybersecurity and data analytics roles. Here in Canada we now have our first federal Minister of AI (11), and in the US their government is pushing to be the leaders in AI (12)
But this narrative is increasingly complicated by the environmental realities. AI data centres are massive consumers of energy and water, two resources already under stress in many regions due to climate change and population growth. In places like Kamloops and Merritt, where power grids are strained and water resources are finite, the introduction of new data centres could exacerbate existing shortages. These facilities often require tens of thousands, if not hundreds of thousands, of gallons of water per day for cooling and can significantly increase regional electricity demand, sometimes drawing from grids that are already overtaxed.
This sets up a difficult question, “are we willing to risk long-term environmental stability and resource availability for short-term job creation?” Or, can we find a more balanced path that doesn’t force communities to choose between sustainability and economic survival? As AI adoption accelerates and the need for computational power grows, these questions will only become more urgent. Communities, policymakers, and families will need to weigh not just the benefits of these new technologies, but also their broader ecological footprint and the trade-offs they bring to the places we call home.
We often think of AI as floating in “the cloud,” invisible and immaterial. But behind every chatbot, image generator, or learning algorithm are power-hungry machines, water-cooled servers, and real-world tradeoffs. AI is thirsty for fresh water and hungry for electricity. As it becomes more integrated into daily life, it’s not just about asking what AI can do for us, but how it will also be placing increased pressure on our already strained water and power infrastructure, something that many of us are not really aware of - just saying!
Digital Food For Thought
The White Hatter
Facts Not Fear, Facts Not Emotions, Enlighten Not Frighten, Know Tech Not No Tech
References:
