The Hidden Cost of AI: Power-Hungry Data Centers Threaten Climate Goals Amid Tech Boom

The Hidden Cost of AI: Power-Hungry Data Centers Threaten Climate Goals Amid Tech Boom

NEW DELHI —NEW DELHI —From chatbots to hyperrealistic video, generative artificial intelligence is pervasive in 2025, but there is a hidden environmental cost to this digital revolution. There are immediate sustainability concerns because the massive data centers, which serve as the actual "brains" of the AI, use a lot of electricity and fresh water.

It just comes down to scale. According to a 2025 Deloitte estimate, data centers will use 536 terawatt-hours (TWh) in 2025 alone, and this number will only increase. According to 2025 reports from the International Energy Association and Goldman Sachs, artificial intelligence could result in a 160% increase in power demand by 2030. In the U.S., a report from the Center for Biological Diversity warned in October 2025 that this boom threatens national climate goals, projecting data centers could claim over 12% of all U.S. electricity by 2030.

This energy drain comes from two parts. First, "training" AI, which is incredibly intensive. Second, "inference," or the energy used for every query. Data from the IEA highlights this spike: a single generative AI query can use 2.9 watt-hours of electricity, nearly ten times the 0.3 watt-hours of a simple Google search.

To meet this demand, tech giants like Microsoft, Google and Amazon are set to spend a combined $375

billion in 2025 alone on building these new, power-hungry facilities. Beyond electricity, AI is incredibly

thirsty. A September 2025 report from Morgan Stanley warned that AI data centers could drive an 11-fold

increase in global water consumption by 2028, reaching over 1,000 billion liters annually. This is reflected at the user level, where a single conversation of 10-50 queries with an AI can consume half a liter of fresh water—the amount in a standard water bottle. According to a 2025 Ecolab study, almost half of American consumers are not even aware that artificial intelligence uses water, despite corporate reports from firms like Microsoft and Google previously demonstrating 20–30% increases in water usage.

As a report of 2025, the Grantham research institute has estimated that AI can cut global emissions by 3.2 to 5.4 billion tonnes by 2035. To make sustainable AI databases, new technologies are emerging. China recently announced an underwater data center that uses sea water for cooling, this way they have already developed a ‘smart grid’ that regulates renewable energy