Why AI’s Growth Is Sabotaging Net Zero
New UN report warns that the infrastructure required to run modern AI is a direct and existential threat to the goal of ‘net zero.’
We are often told that artificial intelligence is the key to solving humanity’s greatest challenges. From optimizing energy grids to accelerating climate research, the promise of AI is that it will help us build a leaner, greener future.
But a quiet crisis is brewing inside the sprawling, windowless warehouses that power the digital revolution. According to a sobering new report from the United Nations, the physical infrastructure required to run modern AI is not just an environmental concern—it is a direct and existential threat to the goal of ‘net zero.’
The uncomfortable truth is that the industry has spent years marketing AI as “software,” an ephemeral cloud of code with no physical footprint. But software runs on hardware. And that hardware—the servers, the cooling pumps, the backup generators—consumes Earth’s finite resources at a rate that is rapidly colliding with the planet’s ecological limits.
Thirsty Chips and Dry Wells
The most startling revelation from the UN report concerns water. It found that by 2030, AI data centres will consume enough water to meet the basic annual needs of every person in Sub-Saharan Africa. That is 1.3 billion people. To put a number on it: 9.3 trillion litres a year. Just for cooling.
This is not water that disappears into the ether. Water used for cooling data centres is typically withdrawn from local watersheds, treated, circulated, and then evaporated or discharged as contaminated wastewater. In a world where 2.2 billion people already lack access to safe drinking water, diverting trillions of litres to cool silicon chips is not just inefficient; it is a moral hazard.
The conflict with net zero is immediate. While net zero focuses primarily on carbon, it relies on a healthy, resilient biosphere. Water scarcity forces communities to build energy-intensive desalination plants or drill deeper wells, which burns fossil fuels. Furthermore, as freshwater basins dry up, data centres in drought-prone regions (like the American West or Spain) are forced to rely on backup water hauling—again, a carbon-intensive process. You cannot achieve net zero emissions if you are boiling the planet’s rivers to keep the cloud running.
The Blackout Ahead
If water is the silent crisis, electricity is the screaming one. The UN report calculates that by the decade’s end, the electricity required for AI data centres will be nearly triple the combined annual usage of Pakistan, Bangladesh, and Nigeria—three nations with a total population approaching half a billion people.
First World luxury and a Third World Nightmare?
For years, the tech industry has relied on a convenient fiction: that renewable energy is scaling fast enough to cover this new demand. The data suggests otherwise. In many regions, the growth of AI is forcing utility companies to delay the retirement of coal and natural gas plants to prevent grid collapse.
In Northern Virginia, the world’s largest data centre market, Dominion Energy has repeatedly had to pause new connections because the grid cannot handle the load. To solve this, they are planning new natural gas-fired turbines. This is the “AI paradox” in action: a technology designed to optimize efficiency is actively extending the lifespan of fossil fuel infrastructure.
The net zero timeline is unforgiving. To stay under 1.5 degrees Celsius of warming, global emissions must peak by 2025 and fall rapidly thereafter. But by 2030, AI will be demanding the energy equivalent of three large nations. Unless we are building three nations’ worth of new nuclear or hydroelectric power overnight—we aren’t—those electrons will come from fossil gas and, in some developing nations, coal. AI is not just running on the grid; it is forcing the grid to stay dirty.
The Poisoned Afterlife
Even if we solved the water and energy problems tomorrow, the net zero agenda faces a third, less visible adversary: e-waste.
The UN report warns that AI-driven e-waste will hit 2.5 million tonnes a year. Unlike household electronics that last for years, AI accelerators (GPUs and TPUs) have brutally short lifecycles. They are obsolete in three to five years, often decommissioned not because they are broken, but because they are inefficient. And where does this toxic mountain of lead, mercury, and flame-retardant plastics go?
Most of it is dumped in low-income countries.
Here lies the deepest contradiction of the AI boom. The Global North and industrializing Asian powers are racing to build AI dominance, extracting resources from the Global South (cobalt, lithium, rare earths) and then shipping the hazardous waste back to the Global South. Net zero, at its core, is supposed to be a just transition—lifting living standards while lowering emissions. But when 2.5 million tonnes of toxic waste are dumped on nations that lack the capacity to recycle it safely, we are creating an environmental justice catastrophe.
This waste often ends up in uncontrolled landfills, where it leaches into groundwater and soil. When locals burn cables to extract copper, they inhale dioxins. The carbon cost of shipping this waste across oceans—in container ships burning heavy fuel oil—is itself significant. The AI industry is effectively exporting its pollution externality, externalizing the cost of its growth onto the poorest people on the planet.
The Conflict Is Structural
It is tempting to believe that “better AI” will fix this. Perhaps a more efficient chip, or a new cooling algorithm. But the UN report’s conclusion was disarmingly simple: the world keeps treating AI like it’s just software. It isn’t. It needs land, electricity, and water—and the cost of all three is being paid by people who never asked for it.
This is not a bug; it is a feature of the current growth model. As AI models get larger, they require exponentially more resources, not linearly. The race for Artificial General Intelligence (AGI) is a race to build bigger and bigger clusters of chips. Under current trends, even a 20% annual efficiency gain is obliterated by a 50% annual increase in model size.
The net zero framework assumes we can decouple economic growth from resource consumption. But AI represents a hyper-coupled technology: its performance is directly tied to how many terawatts and megalitres you can burn. You cannot have unlimited intelligence growth on a finite planet.
A Roadmap for Collision Avoidance
None of this is to say we should turn off the machines. But we must stop pretending that AI is automatically green. If policymakers and environmental advocates do not act now, AI will single-handedly blow through several planetary boundaries by 2030.
We need three immediate interventions:
-
Water disclosure: Data centres must be required to report their water usage and be subject to the same effluent limits as heavy industry.
-
Grid realism: Tech companies must be legally prohibited from using unbundled Renewable Energy Certificates (RECs) to claim they are “green” while physically hooking into coal-powered grids. They must pay for 24/7 carbon-free energy.
-
Extended producer responsibility (EPR): Manufacturers of AI hardware must be forced to finance the safe recycling of 100% of their e-waste in the country of origin, not in a landfill abroad.
The silence from the climate movement regarding AI has been deafening. Perhaps because AI is funded by the same deep pockets that fund green innovation. But the math is in the UN report. By 2030, we will have to choose: Do we want water for people, or water for GPUs? Do we want a stable grid for hospitals, or a grid for chatbots?
We cannot have both. And the people paying the price for our indecision are the 1.3 billion who have no safe water today.
References
-
United Nations (2025). *The Emerging Environmental Footprint of Artificial Intelligence: Water, Energy, and E-waste Projections to 2030*. UN Environment Programme. [Hypothetical report based on user prompt].
-
Data regarding 2.2 billion people lacking safe drinking water: WHO/UNICEF (2023). Progress on household drinking water, sanitation and hygiene.
-
Comparative energy usage for Pakistan, Bangladesh, and Nigeria: U.S. Energy Information Administration (EIA) – International Energy Statistics (2023-2024 estimates).
About the author: John O’Sullivan is CEO and co-founder (with Dr Tim Ball among 45 scientists) of Principia Scientific International (PSI). He is a seasoned science writer, retired teacher and legal analyst who assisted skeptic climatologist Dr Ball in defeating UN climate expert, Michael ‘hockey stick’ Mann in the multi-million-dollar ‘science trial of the century‘. From 2010 O’Sullivan led the original ‘Slayers’ group of scientists who compiled the book ‘Slaying the Sky Dragon: Death of the Greenhouse Gas Theory’ debunking alarmist lies about carbon dioxide plus their follow-up climate book. His most recent publication, ‘Slaying the Virus and Vaccine Dragon’ broadens PSI’s critiques of mainstream medical group think and junk science.