According to Reuters, AI data centers in North America alone will use approximately 1 trillion liters of water by 2025. This amount is roughly equivalent to the annual water demand of New York City. Moreover, the problem isn't just total consumption. The sustainability reports published by companies are often incomplete, fragmented, and inadequately broken down by site. Google reports on its owned and leased sites but may exclude third-party operated sites. Amazon and Microsoft, while disclosing total water usage, don't break it down by site. Therefore, the phrase "green technology" sometimes describes not a real improvement, but a well-packaged ambiguity. This is where the most unsettling aspect of the AI age begins. The industry always describes itself through its results: faster search, more efficient production, smarter software, lower costs, stronger competition.
However, AI requires water, electricity, land, cooling, and transmission lines. And while all this infrastructure is growing, the story told to the public is still almost entirely an abstract tale of progress. Technology companies are selling the future; But those extremely mundane costs of the future—like municipal water, local grids, and regional resource pressures—often pale in comparison.
The electricity side is even more striking. According to the International Energy Agency, data centers consumed 415 TWh in 2024, roughly 1.5 percent of the world's electricity. In the same agency's baseline scenario, this consumption is expected to rise to approximately 945 TWh by 2030, roughly doubling. In other words, artificial intelligence doesn't just demand digital capacity; it demands a massive energy regime.
This demand has accelerated so much that Microsoft is working with Chevron and Engine No. 1 on natural gas-based energy supply to fuel its growing AI data centers; a Bloomberg report states that the proposed facility in Texas has a target production capacity of 2,500 megawatts.
This is where the moral narrative constructed by technology companies collides with the material reality. According to an investor initiative cited in a Reuters report, Alphabet pledged in 2020 to halve its emissions by 2030; however, emissions have increased by 51 percent. On the other hand, Google's 2025 Environmental Report states that the company reduced data center energy emissions by 12 percent in 2024 and replaced 4.5 billion gallons of water, increasing the replacement rate for freshwater consumption from 18 percent to 64 percent. So the defense is ready: efficiency is increasing, innovation is coming, compensation mechanisms are being established. But this defense cannot invalidate the real question: As the overall system continues to grow, who will clearly demonstrate whether the language of improvement is actually reducing the absolute environmental burden? Artificial intelligence is described in terms of progress serving humanity, but it often views nature only as a logistical input.
An analysis published by Reuters on April 1st argues that the AI sector has a systemic blindness to nature, and that its growth model structurally fuels habitat destruction through data centers, energy use, and mineral extraction. This is a very serious but accurate warning. Because today, AI discussions talk about job losses, security, disinformation, and geopolitics; but they don't discuss watersheds, biodiversity, land pressure, and energy resources with the same enthusiasm. The promise of benefit to humanity is becoming a new shield of legitimacy that covers up the cost to nature. Therefore, it is no longer sufficient to read the issue of artificial intelligence solely under the heading of technology. We are faced with a model: first it promises the future, then it expands the infrastructure, and then it relegates the water, electricity, land, and emissions it consumes to footnotes. The real issue is not how intelligent the machines have become, but what that intelligence costs nature.
