AI’s promise is undeniable, but its unchecked expansion risks creating a new industrial footprint as heavy as the fossil fuel era. The challenge for policymakers, companies, and citizens is clear: balance innovation with sustainability, and ensure that the pursuit of intelligence does not come at the expense of the planet’s survival.

Artificial intelligence has become the darling of the digital age. From chatbots that mimic human conversation to image generators that conjure entire worlds, the technology dazzles with its promise. But behind the curtain of innovation lies a sobering truth: AI is devouring the planet’s resources at a pace that should alarm us all.

The Hidden Footprint

• Every AI query requires immense computational power, housed in massive data centers consuming huge amounts of electricity, water, and land.
• Global investment in data centers projected to reach $620 billion by 2026, nearly quadruple 2023 levels.
• GPUs like Nvidia’s A100 rely on over 20 metals, rare earths, and water-intensive semiconductor processes.

Every query to an AI model triggers immense computational power. That power is housed in sprawling data centers, whose construction and operation demand staggering amounts of electricity, water, and land. By 2026, global investment in these facilities is expected to hit $620 billion — nearly four times the level of 2023. Some projects rival the footprint of Manhattan, while their energy draw equals nuclear reactors.

The GPUs at the heart of AI systems are themselves resource-intensive marvels. A single Nvidia A100 chip contains more than 20 metals, including rare earths. Manufacturing wafers for these chips requires thousands of liters of ultra-pure water, alongside toxic chemicals. The semiconductor industry, already notorious for its environmental toll, is being supercharged by AI’s insatiable demand.

Water and Energy Strain

In 2023 alone, data centers withdrew 5,000 billion liters of water — equivalent to France’s annual potable water use. Much of this is lost to evaporation. By 2030, consumption could double. Meanwhile, electricity demand is skyrocketing: U.S. data centers may consume up to 12% of national electricity by 2028, while Europe’s share could triple to 7.5% by 2035.

Despite corporate pledges of sustainability, more than half of the electricity powering these centers still comes from fossil fuels. In 2024, data centers emitted 369 million tonnes of CO₂, more than the entire nation of France. This trajectory risks locking AI into the same carbon-heavy legacy as the fossil fuel industry it claims to disrupt.

A Call for Sobriety

The French ecological agency Ademe has urged “sobriety” in AI development — a reminder that not every problem requires the most resource-hungry solution. In the U.S., more than 230 NGOs have called for a moratorium on new data centers, warning of conflicts with agriculture, transport electrification, and other essential needs.

AI is not inherently destructive. It can help optimize energy grids, accelerate medical research, and model climate change. But the current trajectory — a race toward ever-larger models and ever-bigger data centers — risks undermining those very goals. The question is not whether AI should exist, but whether it should expand without limits.

If the digital revolution is to be remembered as progress rather than plunder, we must demand accountability from the companies driving it. Innovation should serve humanity, not consume its future.

Other People are Reading