Sucked In. The Gaping Maw That Feeds AI Mania

“While technological successes are celebrated, the social fabric is progressively eroded, as if by a silent virus.” — Pope Leo XIV

The artificial intelligence industry likes to refer to its massive “hyperscale” data centres as “campuses.” That’s complete bullshit. A data centre hosts no students, no laughter and no libraries. The poet William Blake would have had a proper name for these ugly bunkers: “dark Satanic Mills.”

Into this mighty gyre now churns $3 trillion of global capital, much of it flowing from debt markets, private credit and government programs that uncritically regard the dangerously flawed technology as nothing short of miraculous, inevitable and necessary. That’s trillions of dollars not being spent to address the cost-of-living crisis or the storms of climate change.

AI’s capital intensity, largely directed by and for a secretive cartel of billionaires, also cultivates a job desert. According to one estimate, it takes a capital investment of $54 million to create one permanent job in the data centre industry while other sectors can create one full-time job with an investment of $322,000. Not to mention the hundreds of millions of jobs AI promises to make obsolete.

The gyre sucks up not only capital but immense supplies of energy. More than 11,000 data centres now occupy the planet, and half are located in the United States. A single 100-megawatt hyperscale facility requires enough electrical juice to light up 100,000 households. A larger proposed AI factory in New Zealand will draw 280 megawatts, or six per cent of that country’s electrical demand. And that facility’s electricity consumption will be dwarfed fivefold and more by behemoths slated for construction around the world.

Already, in Ireland, the data centre industry now consumes 22 per cent of that island’s electricity. That’s more than all of that country’s urban households. The AI juggernaut seems able to manufacture silence by public servants. A recent Irish government report on the importance of data centres just happened to omit any reference to the AI surge and how it pushes up everyone’s electricity bills, the highest in Europe.

Now consider the energy appetite of Kevin O’Leary’s Wonder Valley project in Alberta. O’Leary proposes to spend $70 billion of other people’s money on the planet’s “largest” data centre. It would consume more electricity than is used by eight million households or 15 cities the size of Calgary. Its ravenous energy appetite will tie the robotic entity directly to Western Canada’s Montney formation, its fracked methane used to fire turbines powering the colossus.

If history is a guide, those turbines will foul the air by pumping out lung-clogging fine particulate matter and hazardous chemicals that lower the lifespans of rural people. Alberta Premier Danielle Smith’s government has already waived any pretence of an environmental review.

Hungry for land and water

In addition to swallowing capital and energy, the gyre also absorbs forests, grasslands, suburbia and farms into its devastating spiral. Data centre footprints keep expanding along with demand for computing power so that now a hyperscale factory typically sprawls over 500 acres — 3 1/2 times the size of Disneyland. But what if regulators, wanting to “green” data centres, stipulate they be powered by utility-scale solar or wind power? That saves methane but affects even more land.

There seem to be no limits in sight. One Montana project proposes to plant an edifice the size of 3,000 football fields in the middle of cattle country. The project will be 31 times larger than its closest rural community, Broadview. O’Leary’s project would similarly fill some 8,000 acres of parkland south of Grande Prairie. That’s a swath of Alberta larger than Manhattan.

Into this gyre of land loss churn mountains of metal and minerals requiring extensive mining and processing. Copper and aluminum for power and cooling systems. Lithium or diesel for backup batteries. Gallium and germanium for the semiconductors. According to the World Economic Forum, every megawatt consumed by a data centre requires 60 to 75 tonnes of materials mainly for its power and cooling systems. As the data bunkers get larger and spend more energy, this material intensity grows like a city.

AI advocates will say to relax, because all this mind-boggling consumption nourishes the stock market and drives the hallowed metric of GDP. But most resources cannot be endlessly spun into stock plays. There are limits, for example, to clean, fresh water. The AI gyre depends on great volumes of water and is already stressing out drought-prone regions throughout North America. By 2028 data centres will consume enough fresh water in the United States to quench the indoor water needs of 18 million people.

AI actually begins as a water hog. The manufacturing of semiconductor chips, which demands ultra-pure water for fabrication, involves 400 chemicals including cancer-causing per- and polyfluoroalkyl substances that are virtually indestructible in nature. Complex chip production, researchers say, discharges a toxic brew of “many unknown and hazardous contaminants.”

When the finished chips are assembled on racks in a data centre, the warehouse’s cooling system requires even more water to keep the facility from overheating.

According to J.P. Morgan, a large data centre can consume as much water as a chip manufacturer: about 19 million litres per day. That’s the same amount of water used by a town of up to 50,000 people. Every internet search aided by AI comes soaked in fresh water, about a bottle’s worth of the world’s most critical resource.

Forced onto ‘heat islands’

As the gyre consumes and digests, it expels heat. Enough to pose a threat to humans and nature at its pulling edges.

It’s common knowledge that by replacing soil and trees with concrete and asphalt, the modern megacity has created what scientists call “urban heat islands.” When temperatures hit 29 C in Sydney, Australia, the city’s urban core actually records temperatures four degrees higher. In Paris the urban heat island effect can increase the average temperature by two to three degrees in the summer.

The capture of this heat has the perverse effect of burning more fossil fuels at power plants to keep overheated people cool via air conditioners, which, in turn, creates more volatile weather, greenhouse gas emissions and air pollution.

Now the world has a new contributor to the heat island effect. Data centres are raising the temperature of nearby communities and posing additional threats to their health.

A recent study by scientists from the United Kingdom, Singapore, France and Hong Kong disclosed that data centres dissipate so much heat while performing AI calculations that they, in effect, create their own microclimate zone. The researchers called this “the data heat island effect.”

Using remote sensing data from satellites, the researchers located 6,000 data centres built in the last decade in rural regions and measured temperature changes before and after construction. Everywhere they found that land surface temperatures had increased by an average of 2 C. In one extreme case temperatures soared by 9 C.

The researchers discovered that the hotter temperatures extended up to 10 kilometres’ distance from the AI hyperscaled facilities. Many of the places they tracked are already plenty warm. In the northeast region of Brazil, for example, where during the hottest months daily highs hover in the high 30s, data centres have bumped local temperatures another nearly 3 C.

Pulling back their lens, the scientists considered how many global inhabitants live within a 10-kilometre radius of a big AI data centre. Their calculation: about 343 million people. Their conclusion: “Especially in the context of global warming and climate transformation,” the data heat island effect may lead “to dramatic impact on welfare, health care and energy systems.”

The mounting resistance

Regardless, the gyre spins with rising speed and force. Unless it is somehow made to slow, the next five years will see data centre power consumption exceed the amount used by all the world’s global manufacturing industries. According to the United Nations, the water demands of data centres by 2030 could be more than nine trillion litres. That’s enough water to quench the thirsts of all 1.3 billion people in sub-Saharan Africa.

Canadians can be forgiven for feeling as if this vortex of destruction is being sold to them at the rushed clip of an unintelligible Albertan auctioneer. To date, nearly 100 supersize data centres have been proposed for construction in Canada. About 90 per cent of these projects will be built in rural Alberta, where the government has slavishly courted the industry without any democratic mandate to do so. Or transparency about the ecological costs.

A recent York University paper correctly noted that the rapid expansion of supersize data factories represents “a structural shift rather than incremental growth.” It warned that Alberta’s data centre ambitions threaten to challenge grid reliability, increase methane pollution and totally unsettle rural communities. “The result is a concentration of planned capacity in a moderate-to-high water risk province heavily dependent on natural gas whose grid emissions intensity is nearly five times the national average.”

So the AI gyre gobbles capital, land and energy. It slurps scarce water. It accelerates the mining of metals and minerals amid a growing geopolitical disorder exacerbated by resource scarcity. It concentrates economic power in fewer hands, and largely foreign ones at that. And it enables a totalitarian infrastructure designed to accelerate the automation of everything and progressively erode the human fabric.

Worry not. The federal government wants to provide “AI literacy training” to those of us not already completely sucked in.

But resistance does grow. When residents of Olds, Alberta, learned Canada’s largest AI data centre was slated for their small farming community, they organized their own educational sessions and raised alarms. The province’s regulator halted the project, though its backers have reapplied.

“These projects are being pushed on us with no accountability, while the communities are left with the consequences,” said Jesse Cardinal, executive director of the Indigenous-led group Keepers of the Water. He could have been speaking for every person, everywhere, waking up to the ruthlessly marketed threat.

As Google’s own AI will tell you: “Public resistance to data centres is mounting rapidly, with recent polls showing majorities — and even working-class neighbourhoods — strongly opposing nearby builds.” According to Data Center Watch, local opponents have blocked or delayed U.S. data centre projects worth $64 billion.

Those citizens are at the forefront of a larger fight that pits human dignity and reason against the swirling void that is the AI obsession.