{
  "id": 16175667,
  "title": "Big Tech’s thirst for water is sparking a global resistance",
  "url": "https://urgent.news/2026/10/23/big-techs-thirst-for-water-is-sparking-a-global-resistance",
  "topic": "world",
  "section": "World",
  "published": "2026-10-23T11:07:32.000Z",
  "source": {
    "name": "Rest of World",
    "slug": "restofworld",
    "url": "https://elink32b.restofworld.org/ss/c/u001.wqTatvCe1uEimPApFRU3vFduVjiFuG_K8CXAMZILUNve8PULprmp55qy0pVDkT-Bdgl9kXUrXfFQ0mIHLoyJ5DSwOy2jGtbTlOYW5Aj_3-g_CXRr8gOtocUuF966XiD4b3mdVPtMNhrIqWstmthzhQVwWVBFLNPH-guJaBFgwLS9V59F5P4ix-a2RzlPrkBblL-Iw9NGKBQn_g3CuBDtuWyW-RxnOuFG_0toyR0kSVT8K7cVZPmptMQhenGGhLURw-DD7q9NGHrtx_IuAMavUwNbw9hQKp6t8S1cLoTi889gzia0klDRujclqLqiExsS6qCn5qsmuWvA7nKkJx9VDg/4v4/mDnpXwTcRKKBZklsXZpUIg/h1/h001.ANUF1uR5_AhRTpgTu1-T3kk1A_LxKXh3z6XQz5dX7jI"
  },
  "original_language": "en",
  "account": "In November 2022, the tech industry celebrated the release of ChatGPT as a symbol of technological progress. However, the water demands of these ambitious projects remained shrouded in mystery. A 2023 University of California study estimated that training the AI model behind ChatGPT required 185,000 gallons of water. Each interaction with the chatbot or request for an email draft seemed to equate to pouring a bottle of water down the drain. This revelation sparked public concern, prompting tech giants like Google and OpenAI to provide their own water usage figures. These companies claimed that querying their chatbots used minimal amounts of water, but independent researchers dismissed these assertions as overly optimistic. Data centers, which power these AI models, consume vast amounts of water, with some using between 1 and 5 million gallons daily—comparable to the water usage of a small city. As these centers grow, so do their water needs. GPUs, crucial to generative AI, are far more resource-intensive and heat-generating than traditional data center components, necessitating even more water for cooling. Water sources are a critical concern for these facilities, often drawing from public drinking water supplies and competing with local communities. Chile serves as a prime example of this issue. In October 2019, nationwide protests erupted, with residents rallying against the proposed placement of a $200 million data center in Cerrillos, Santiago. Google marketed the project as a boon, promising jobs and prosperity, yet it would consume an astounding 169 liters of water per second—equivalent to all residents of Cerrillos using drinking water for a year. To make matters worse, the data center aimed to extract water from an aquifer under a 2005 decree that prohibited new water extraction. Just weeks before the project received approval, this decree was revoked, leaving local activists to fight against Google's plans. Despite months of advocacy, a December 2019 referendum saw 49% of Cerrillos residents reject the project, while 38% supported it. Following a lawsuit in 2020, Google proposed cooling the data center with air conditioning instead of water. Similarly, in Quilicura, opposition grew as Microsoft planned to build a data center in 2022 amid strict water rationing. The struggle for water resources in Chile highlights the growing tension between large tech companies and local communities over water usage.",
  "summary": null,
  "key_points": [
    "ChatGPT's training required 185,000 gallons of water in 2023.",
    "Data centers power AI models, consuming 1-5 million gallons of water daily.",
    "Chile protests against data centers consuming 169 liters of water per second."
  ],
  "editors_take": "The massive cooling requirements of generative AI infrastructure force a direct confrontation between global technology giants and local communities competing for finite public drinking water supplies.",
  "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."
}