A recent study assessed the environmental impact of artificial intelligence in 2025 and called on tech companies to be more transparent about their polluting emissions and water usage. According to the latest estimates, the carbon emissions generated by artificial intelligence this year are equivalent to New York City's overall emissions level, and the amount of water consumed by people around the world through bottled water is equivalent to the total amount of water consumed by people around the world. “It's impossible to come up with an extremely accurate number, but in any case, the scale will be huge... Ultimately, everyone is paying for it.” Alex de Vries-Gao, a doctoral student at the Environmental Research Institute at the Vrije University of Amsterdam, said. He published the study in the journal Patterns. “At the end of the day, everyone is paying for it.” To calculate these data, Alex de Vries-Gao made estimates based on previous research. Early research suggests that global AI's electricity demand could reach 23 gigawatts this year, exceeding the electricity consumed by Bitcoin mining in 2024. Although many tech companies disclose overall carbon emissions and direct water consumption in their annual sustainability reports, they generally don't break down how many resources artificial intelligence consumes. Alex de Vries-Gao indirectly estimates the production scale of artificial intelligence-related hardware and the level of energy these hardware may consume through analyst estimates, company earnings conference calls, and other public information.

A recent study assessed the environmental impact of artificial intelligence in 2025 and called on tech companies to be more transparent about their polluting emissions and water usage. According to the latest estimates, the carbon emissions generated by artificial intelligence this year are equivalent to New York City's overall emissions level, and the amount of water consumed by people around the world through bottled water is equivalent to the total amount of water consumed by people around the world. “It's impossible to come up with an extremely accurate number, but in any case, the scale will be huge... Ultimately, everyone is paying for it.” Alex de Vries-Gao, a doctoral student at the Environmental Research Institute at the Vrije University of Amsterdam, said. He published the study in the journal Patterns. “At the end of the day, everyone is paying for it.” To calculate these data, Alex de Vries-Gao made estimates based on previous research. Early research suggests that global AI's electricity demand could reach 23 gigawatts this year, exceeding the electricity consumed by Bitcoin mining in 2024. Although many tech companies disclose overall carbon emissions and direct water consumption in their annual sustainability reports, they generally don't break down how many resources artificial intelligence consumes. Alex de Vries-Gao indirectly estimates the production scale of artificial intelligence-related hardware and the level of energy these hardware may consume through analyst estimates, company earnings conference calls, and other public information.