{
  "id": 6605508,
  "title": "AI is thirsty: Fresh water has become a key resource of the digital economy",
  "url": "https://urgent.news/2026/09/10/ai-is-thirsty-fresh-water-has-become-a-key-resource-of-the-digital",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-10T14:39:58.000Z",
  "source": {
    "name": "The Insider",
    "slug": "the-insider",
    "url": "https://theins.press/en/economics/296921"
  },
  "original_language": "en",
  "account": "Artificial intelligence (AI) is rapidly consuming vast amounts of fresh water to power data centers, according to emerging research. These massive industrial facilities house servers that continuously train and run neural networks, requiring large quantities of clean water for power generation and cooling. It is projected that AI's fresh water consumption will double between 2025 and 2030, reaching around 9 trillion liters per year - equivalent to the annual water consumption of countries like Greece or Poland.\n\nThe water footprint of AI varies depending on the complexity of queries. A single Google Gemini search query consumes about 2.37 ml of water, while generating a short video requires more than 4 liters. As engineers improve algorithms and cooling systems, the cost of individual queries falls, leading to an explosive increase in AI adoption. This creates a Jevons paradox, where efficiency gains in resource use do not reduce consumption but instead increase it.\n\nAI's water footprint can be divided into three interconnected levels: direct water consumption at data centers, indirect water consumption in electricity generation, and the entire supply chain. Direct consumption is primarily due to evaporative cooling, which typically results in 80% of water drawn being lost as vapor. Indirect consumption comes from traditional thermal power plants that require significant amounts of fresh water to cool steam turbines. The manufacturing of high-performance microchips also necessitates thousands of liters of ultrapure water for rinsing silicon wafers at every stage of production.\n\nEfforts to make AI greener through advanced closed-loop water treatment systems and alternative cooling methods have unintended consequences, further increasing water consumption. Cold-plate cooling and single-phase immersion cooling can reduce water use by up to 52%, but these technologies require substantial capital investment and changes to server architecture. Additionally, the use of fluorinated PFAS compounds in some cooling systems poses environmental risks, as these \"forever chemicals\" are highly persistent and toxic.",
  "summary": "Lobbyists for Microsoft, Amazon, Google, and Meta have succeeded in getting the European Commission to keep data on data center water consumption secret, European investigative journalists have found . The huge amounts of fresh water needed to cool servers are fueling local conflicts over water resources amid a global drought. Protests against data center construction are breaking out around the…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}