AI Data Centres in Canada: Why the Sudden Boom?

The hum of progress has a new, unmistakable sound in Canada: the low, constant whir of cooling fans and high-voltage transformers. In 2025, the nation’s ambition to become a global AI leader is no longer measured just in algorithms or venture capital, but in concrete, steel, and staggering resource demands. A surge of proposed and under-construction AI data centres is reshaping landscapes and sparking fierce community debates, forcing a pressing question: what is the true cost of housing the brains of our digital future?

Unlike the server rooms of the past, these are not mere digital warehouses. They are industrial-scale power plants for computation. “The facilities we’re building for AI are just bigger and bigger,” explains Anne Pasek, Canada Research Chair in Media, Culture and the Environment at Trent University. The shift is fundamental. Traditional cloud facilities might draw a couple of megawatts. The new AI hyperscale centres? They start at fifty and routinely exceed one hundred. To put that in perspective, just 50 megawatts can power 37,500 homes. A recent York University study paints a startling picture of scale: while Canada currently has only five hyperscale centres, 159 facilities were proposed or under construction as of July, with 116 of those expected to surpass 100 megawatts—a first for the country.

This boom is a direct consequence of national strategy meeting global frenzy. With Prime Minister Mark Carney’s “AI for All” strategy pledging billions and aiming to catapult business adoption from 12% to 60% by 2034, the signal to developers is clear. Proponents like Martin Karsten, a computer science professor at the University of Waterloo, frame it as an inevitable facet of modern life. “As long as we are all very interested in using these [AI] services,” he notes, “I think we have to accept the fact that these data centres keep coming.” The economic promise is a powerful lure, especially for provinces like Alberta, which could see its share of national data centre capacity explode from 3% to 93% if all proposed projects proceed.

But acceptance is far from universal. The sheer physical footprint of these projects—some the size of multiple football fields—is only the beginning of community concerns. The real friction points are water and watts. AI’s powerful GPU chips generate immense heat, requiring massive cooling. Many systems use evaporative cooling, pulling vast quantities of water from local watersheds only for it to vanish as vapor, effectively exporting a region’s water elsewhere. A single 100-megawatt facility can consume about 2.5 billion litres annually, enough for 80,000 people. In an era where, as Kaveh Madani of the UN University Institute for Water, Environment and Health states, “We are already water bankrupt in many places,” this new demand is a profound stress test.

The energy appetite is equally disruptive. Data centres accounted for 4.4% of U.S. electricity demand in 2023, a figure projected to potentially triple by 2028. In Canada, the strain threatens grid stability during peak events like heatwaves and could drive up consumer power bills. Furthermore, because connecting to the grid can take years, many new centres are being designed with on-site power plants fueled by natural gas. This introduces direct health risks to nearby communities. Fossil-fuel combustion emits fine particulate matter (PM2.5), a pollutant linked by extensive research to increased risks of respiratory, cardiovascular, and neurocognitive problems. An independent analysis of a Virginia AI centre found its natural gas turbines and diesel backups could cause tens of millions in annual health damages.

Then there’s the audible presence: a 24/7 industrial symphony of fans, generators, and a pervasive low-frequency hum. Chronic noise pollution is a well-documented source of stress, sleep disturbance, and long-term health issues. It’s a tangible grievance, with residents in over a dozen U.S. states filing lawsuits in recent years. The combination of these impacts has led communities from Oakville to Mississauga to hit pause, enacting moratoriums to demand more answers.

The response from experts isn’t a simple “yes” or “no” to development, but a call for smarter, more equitable design. The trade-offs are clear:

• Optimally efficient facilities powered with renewable energy
• Using closed-loop cooling systems that recycle water
• Managing noise through better engineering and zoning
• Locating centres remotely to reduce population exposure
• Consulting rural and Indigenous communities
• Rethinking the scale of ambition itself

Perhaps the most provocative alternative is to rethink the scale of ambition itself. Pasek suggests spreading computing capacity more broadly to “break up the burden” or even pivoting toward smaller, open-source AI development models that don’t require hyperscale infrastructure to be profitable. It’s a question of values: is our AI future necessarily one of concentrated, gargantuan consumption, or can we architect a more distributed and sustainable path?

The data centre boom is more than a construction trend; it’s a physical manifestation of our digital priorities. As these facilities proliferate, they force a long-overdue conversation about infrastructure, community health, and environmental limits in the AI age. The humming in Canada’s backyard is the sound of the future arriving. The question now is whether we can shape it to be a sound we can live with.