The Machine
Your company just replaced half the content team with AI. Your cousin's call centre closed. A friend in graphic design can't find a contract. Meanwhile the server farms that power the chatbots are drinking lakes of water, burning through electricity grids, and the billionaires building them say don't worry — they'll send you a cheque. This thread is not about whether AI is good or bad. It's about who pays the environmental bill, who loses the paycheque, and whether the proposed safety net is a genuine floor or a gilded leash.
The System
Before the promises or the politics, look at what AI physically requires. A large language model is not a brain in a jar — it is a warehouse of GPUs drawing power around the clock, rejecting heat into cooling towers that evaporate water by the billion litres, and refreshing hardware on cycles short enough to generate mountains of electronic waste. Globally, data centres already consume more electricity than most countries. In Canada, the grid impact is arriving faster than planners expected — Quebec is doubling electricity rates for large data centres, Ontario is gate-keeping new connections, and the Canadian Climate Institute warns that data-centre demand could increase national electricity use by more than 400% by 2050 if left unchecked. The environmental ledger is not hypothetical. It is metered, billed, and growing.
Global data centre electricity (2024)
Google GHG emissions increase since 2019
US data centre water consumption (2023)
Potential Canadian data centre electricity increase by 2050
Global data centre electricity consumption reached approximately 415 terawatt-hours in 2024 — about 1.5% of global electricity use — growing at 12% per year over the previous five years. AI-focused data centres alone grew 50% in 2025 and are projected to triple by 2030.
Google's greenhouse gas emissions rose 48% since 2019, driven by data centre energy demand. Microsoft's Scope 2 emissions more than doubled between 2020 and 2024, rising from 4.3 to nearly 10 million metric tonnes of CO2 — despite a pledge to be carbon negative by 2030.
U.S. data centres consumed 17 billion gallons of water directly through cooling in 2023. Lawrence Berkeley National Lab projects that figure could double or quadruple by 2028. Microsoft's global water consumption rose 34% in a single year to 6.4 million cubic metres. Google's rose to 6.1 billion gallons by 2024.
In Canada, data centre electricity demand could increase by more than 400% by 2050. Quebec is doubling electricity rates for data centres exceeding 5 megawatts starting mid-2026, while Ontario now requires Ministerial approval before large facilities connect to the grid — data centres could account for 13% of Ontario's new electricity demand by 2035.
The Promise
Every automation wave arrives with the same two-part sales pitch. Part one: this will make everyone richer. Part two: anyone displaced will be taken care of. The AI era is no exception. Goldman Sachs says generative AI could raise global GDP by 7%. The IMF says 40% of jobs worldwide are exposed. The World Economic Forum says 85 million jobs will be displaced but 97 million new ones will appear. Tech leaders lined up behind UBI as the solution — Sam Altman spent $14 million on a basic income experiment, Elon Musk endorsed the idea, Andrew Yang ran for president on it. But Altman has since walked back his support, pivoting to 'universal basic compute' and ownership models through Worldcoin. By April 2026, OpenAI itself was proposing robot taxes and a public wealth fund — the company building the displacement machine now writing the policy paper for the safety net. A peer-reviewed paper from Simon Fraser University called this dynamic 'symbolic violence': the tech elite's UBI advocacy normalizes displacement as inevitable, positions the disruptors as benefactors, and preempts structural demands like democratic AI governance. The consistent through-line is that the people building the machine are also proposing the cushion — and they keep redesigning both.
Jobs globally exposed to generative AI (Goldman Sachs)
Canadian workers with significant AI exposure (StatsCan)
Big Tech AI infrastructure spending (2025)
Goldman Sachs estimates generative AI could expose approximately 300 million full-time jobs globally — about 9% of all employment — while potentially raising global GDP by 7% over a decade. The IMF estimates 40% of all jobs worldwide have significant AI exposure, rising to 60% in advanced economies.
The World Economic Forum's 2025 Future of Jobs Report projects that by 2030, 85 million jobs will be displaced worldwide, while 97 million new roles will emerge — for a net gain of 12 million positions. The report notes 40% of employers plan to cut roles where AI handles routine tasks.
Microsoft, Google, Amazon, and Meta spent a combined $320 billion on AI infrastructure in 2025, more than double the $151 billion in 2023 — with projections of $562 billion in 2026. The five largest tech companies by market cap are now worth a combined $21.5 trillion, representing roughly 16% of all global stock value.
Canada's Parliamentary Budget Officer costed a national Guaranteed Basic Income in February 2025: gross payments of $107 billion per year, but a net federal cost of $3.6 to $5 billion after replacing overlapping programs. The PBO estimated it would reduce national poverty by 34-40%.
In April 2026, OpenAI itself released a 13-page policy blueprint proposing robot taxes on companies that replace workers with AI, a public wealth fund giving citizens a stake in AI-driven growth, and a 32-hour workweek pilot. The document acknowledged that AI could hollow out payroll tax revenue funding Social Security, Medicaid, and housing assistance. A peer-reviewed 2025 paper in Frontiers in Artificial Intelligence argued that tech-elite UBI advocacy constitutes 'symbolic violence' — a bid for social license that normalizes displacement as inevitable, positions tech companies as benevolent providers rather than causes of disruption, and preempts more radical demands like democratic control of AI.
The Reality
So what has actually happened? In Canada, Statistics Canada found that AI has not yet caused broad job losses — even in highly exposed occupations — but younger workers and those without degrees saw weaker job growth between late 2022 and late 2025. About 60% of Canadian employees are exposed to AI-related job transformation. Only 12% of Canadian businesses reported using AI by spring 2025, but that had doubled from 6% just one year earlier. Globally, specific companies have already moved: Klarna replaced 700 workers with AI, then admitted quality collapsed and started rehiring. IBM announced 7,800 back-office jobs would go to AI. BT plans to eliminate 55,000 positions by 2030, 10,000 to AI. The deeper structural trend is labour's shrinking share of income — the ILO reports it fell from 53% to 52.4% of global GDP between 2014 and 2024, a gap worth $1 trillion per year. Technology is responsible for roughly half that decline in advanced economies. Meanwhile, the Ontario basic income pilot — the one Canadian experiment that generated hard data — was cancelled after one year for political reasons. The participants who were studied showed 75% kept working, 88% reported less stress, 73% less depression. Then the data tap was turned off. The counterargument deserves a serious hearing. Bank of America points out that 60% of today's jobs didn't exist in 1940 — automation has historically created more work than it destroyed. When ATMs arrived, bank teller employment actually doubled because cheaper branches meant more branches. But the story has a sequel: once smartphones let people bank from home, teller jobs fell off a cliff. The ATM automated a task; the iPhone eliminated the reason the task existed. AI may be closer to the iPhone than the ATM. That question is genuinely open. What is not open is the environmental cost. That is metered. And the wealth concentration. That is audited. The question is what happens to the people in between.
About 60% of Canadian employees are exposed to AI-related job transformation. Roughly half are in roles where AI is highly complementary (AI supports the worker); the other half are in roles where AI can adequately perform key functions. Only 12% of Canadian businesses reported using AI to deliver services by spring 2025, up from 6% a year earlier.
Labour's share of global GDP fell from 53% in 2014 to 52.4% in 2024 — a gap worth $1 trillion per year in lost worker income. In advanced economies, about half of the decline is attributable to technology, the other half to globalization. The share going to capital owners has risen correspondingly.
Ontario's basic income pilot (2018-2019) gave up to $16,989/year per single person. McMaster University research found 75% of recipients who were working kept working, 88% reported less stress and anxiety, 73% less depression, and health system usage dropped. The Conservative government cancelled the program after one year — before the formal evaluation was complete.
Bank of America research (May 2026) notes that 60% of jobs that exist today did not exist in 1940 — historical automation waves have consistently created more work than they destroyed. When ATMs were introduced, bank teller employment doubled as cheaper branches meant more branches. But once smartphones made branch visits unnecessary, teller jobs collapsed. The question is whether AI automates tasks within the current paradigm or eliminates the paradigm itself.
What Works
Three policy levers have evidence behind them — none of them are UBI cheques from tech billionaires. First: direct income floors work when they are government-run and unconditional. Canada's own Ontario pilot showed people kept working and got healthier. Finland's 2017 experiment showed a small employment increase and major wellbeing gains. The PBO says a national guaranteed basic income would cost the federal government $3.6-5 billion net — less than one percent of the AI infrastructure spending Big Tech committed in 2025 alone. Second: grid governance matters. Quebec's move to double data centre electricity rates and Ontario's Ministerial approval requirement for grid connections show that provinces can set terms — not just welcome server farms and hand them the bill. The Canadian Climate Institute argues for integrating data centres into electricity planning rather than treating them as just another customer. Third: the countries that navigated prior automation waves best — Denmark, Sweden, the Netherlands — did not pay people to stay home. They invested in active labour market policy: retraining, placement, wage subsidies during transition. Canada spends 0.2% of GDP on active labour market policies. Denmark spends 1.9%. The gap is a policy choice, not a law of nature.
Canada spends approximately 0.2% of GDP on active labour market policies — training, placement, retraining — compared to the OECD average of 0.5% and Denmark's 1.9%. This is the investment that determines whether displaced workers retrain or fall through the floor.
Quebec's decision to double electricity rates for data centres over 5 MW and Ontario's new Ministerial approval requirement for grid connections demonstrate that provinces can set terms for AI infrastructure rather than subsidizing it. The Canadian Climate Institute recommends integrating data centres into formal electricity planning — treating them as grid participants, not just customers.
Every major institutional projection — Goldman Sachs, WEF, IMF, McKinsey — shows net positive job creation through the AI transition, with 97 million new roles expected against 85 million displaced by 2030. But the new jobs require different skills, arrive in different geographies, and pay different wages. The transition gap — who retrains, who pays, how long it takes — is the policy question that determines whether the net positive is real or statistical fiction.
What You Can Do
If you have read this far, the through-line is clear: AI is not the weather. It is infrastructure — built with permits, powered by public grids, deployed by companies with quarterly earnings. That means it is governable. Follow the water and electricity disclosures your province publishes when data centres apply for grid connections. Ask your MP whether Canada will match the OECD average for retraining investment before the displacement curve steepens. Track the Statistics Canada AI employment series — it updates quarterly and it's the closest thing to a real-time national dashboard on who is gaining and who is losing. And if someone tells you a monthly cheque from a tech billionaire is the solution, ask who controls the off switch.
Subscribe to Statistics Canada's quarterly employment trends series on generative AI — it tracks which occupations are growing and which are contracting in AI-exposed sectors, broken down by age, education, and province.
File a request to your provincial utility or energy regulator for the water and electricity impact assessments filed by data centres seeking grid connections in your region. These are public record in most provinces. When communities know what the server farm costs, the cost-benefit conversation changes.