The Liquid and Land Toll of AI & Data Farms

Part 2 of an on-going series on the impact of these modern technology farms

The proposed 9-gigawatt Greenview AI site in northern Alberta and the identical Stratos AI project now in Utah will each dump the thermal energy equivalent heat signature of twenty-three atomic bombs exploding in that region every single day. Local temperatures will spike by a staggering 15.5⁰ C (28 degrees Fahrenheit).

Let that sink in for a minute!

If you were to repeat the twenty-three atomic bombs statistic to your neighbors, they might think you are exaggerating. But; the math is open, verifiable, and terrifying. According to published analysis by Utah State University physicist Dr. Robert Davies, a fully realized 9-gigawatt hyperscale facility requires an on-site fuel burn of roughly 16 gigawatts to power its computational infrastructure.

Every single watt of power pushed through a computer chip ultimately turns into physical heat, that facility acts as a 16-gigawatt radiator venting directly into the local air and nearby water table twenty-four hours a day. In the language of physics, 16 gigawatts mean the facility is unleashing 16 billion Joules of heat energy every single second. Over the course of a single twenty-four-hour day, that adds up to 1.38 quadrillion Joules of thermal energy dumped into the region.

To put that in perspective, the atomic bomb dropped on Hiroshima released roughly 63 trillion Joules of energy. If you divide the daily thermal output of this data center by the energy of that historic blast, the math leaves no room for corporate spin. The facility dumps the exact thermal equivalent of twenty-two to twenty-three atomic bombs into the local ecosystem every single day, transformed entirely into a silent, invisible thermal wave.

If that doesn't startle you into rethinking your position on AI or Data farms, the numerous massive industrial footprints currently being fast-tracked across North America are sure to shock you even more.

A single proposed megaproject like the Greenview site can require up to 24 million cubic metres of water annually. Let's put that in context.

In a comprehensive investigative report published by The Walrus in May 2026, details emerged regarding "Wonder Valley," the massive 70-billion-dollar AI data centre industrial park backed by celebrity investor Kevin O'Leary and proposed for the Municipal District of Greenview, just south of Grande Prairie, Alberta. The project intends to divert up to 24 million cubic metres of water every year from the Smoky River. This massive draw is slated for a region where the local municipality was forced to declare an agricultural disaster due to severe, prolonged drought conditions.

To cool the tens of thousands of high-density server racks required for generative AI, hyperscalers must rely on immense evaporative cooling systems or massive industrial chillers. When water evaporates to cool a data center, it is completely lost to the local watershed. This poses an immediate threat to agricultural communities that rely on predictable water tables for irrigation pivots and livestock.

One cubic metre of water is exactly 1,000 litres. Therefore, 24 million cubic metres equals 24 billion litres of water. Here is how that volume breaks down using common, real-world benchmarks:

The average Canadian household consumes roughly 300 cubic metres of water per year for all indoor and outdoor needs, including showers, washing machines, toilets, and lawn watering. The Greenview project matches the total, combined annual water usage of 80,000 typical suburban homes. That is equivalent to the entire domestic water demand of Lethbridge and Red Deer combined.

Think of a standard kitchen faucet running completely wide open at a typical household flow rate of about 8 litres per minute. To empty 24 billion litres of water, you would have to turn on 5,700 kitchen taps simultaneously and leave them running completely wide open, 24 hours a day, 365 days a year, without ever turning them off.

At the same time, the Greenview site will swallow 64 sq. km's of land. That is the equivalent of the total land area of Lethbridge or of Red Deer.

These are not hypothetical science fiction scenarios. They are the actual engineering metrics of the global data center rush, and these numbers are fully documented.

The land footprint of these facilities represents a total corporate restructuring of rural geography. At 65 sq. km's (16,000 acres), Alberta's Wonder Valley is not a mere office park. This is a sprawling, windowless concrete monolith that severs wildlife corridors, permanently removes productive agricultural land from local economies, and alters localized weather patterns.

In June 2026, CBC News reported a growing wave of public backlash at community open houses in northern Alberta, where rural residents and agricultural producers expressed deep anger over how these massive footprints were being fast-tracked. The Sturgeon Lake Cree Nation has launched formal legal challenges against the Alberta government, pointing out that the province completely waived the requirement for a provincial environmental impact assessment for the Wonder Valley project, despite the immense risks of localized heat islands and massive water depletion in an area already ravaged by wildfire.

We have been told that the digital economy is a green, weightless alternative to traditional heavy industry. The reality is that a single generative AI text prompt requires a small sip of water, but when multiplied by billions of queries a day across the globe, it transforms into an environmental sledgehammer. When international tech consortia can buy up local water rights, bypass standard environmental reviews, and dump twenty-three atomic bombs worth of heat into rural locations every single day just to power automated chat responses and digital image generation, the local community bears the entire physical burden.

Frequently Asked Questions

Q1. How much water will the Wonder Valley AI data centre use?

The Wonder Valley project in Alberta's Municipal District of Greenview plans to divert up to 24 million cubic metres of water every year from the Smoky River. That equals 24 billion litres, roughly the combined annual water usage of 80,000 typical Canadian suburban homes, in a region already suffering from severe drought.

Q2. How much heat does a 9-gigawatt AI data centre release?

A fully realized 9-gigawatt hyperscale facility requires an on-site fuel burn of roughly 16 gigawatts, and every watt eventually turns into heat. That works out to about 1.38 quadrillion Joules of thermal energy per day, the equivalent of twenty-two to twenty-three Hiroshima-sized atomic bombs released into the local environment every single day.

Q3. Why do AI data farms consume so much water?

Generative AI runs on tens of thousands of high-density server racks that must be cooled constantly. Hyperscalers rely on evaporative cooling systems and industrial chillers, and when water evaporates to cool a data centre, it is permanently lost to the local watershed instead of being returned for farms, livestock, and households.

Q4. How much land do these AI data centre projects occupy?

The Greenview site alone will cover roughly 64 to 65 square kilometres, or about 16,000 acres. That is comparable to the entire land area of Lethbridge or Red Deer, and the footprint severs wildlife corridors, removes productive farmland from local economies, and can alter localized weather patterns.

Keep an eye out for the next installment of this series, where we will examine the atmospheric toll of these projects, investigating the millions of tons of carbon emissions generated by the specialized concrete formulations required to build these planetary-scale computational warehouses