2025’s AI Boom Drives Significant CO2 Emissions and Water Consumption, Study Reveals

Essential brief

Key facts

AI-related carbon emissions in 2025 matched those of New York City, highlighting AI’s significant climate impact.

AI’s water consumption now exceeds global bottled-water demand, indicating substantial resource use.

AI emissions account for over 8% of global aviation emissions, emphasizing its growing environmental footprint.

Current environmental costs of AI are largely borne by society, raising questions about tech companies’ accountability.

Rapid expansion of large-scale data centers driven by AI increases energy consumption and emissions worldwide.

Highlights

AI-related carbon emissions in 2025 matched those of New York City, highlighting AI’s significant climate impact.

AI’s water consumption now exceeds global bottled-water demand, indicating substantial resource use.

AI emissions account for over 8% of global aviation emissions, emphasizing its growing environmental footprint.

Current environmental costs of AI are largely borne by society, raising questions about tech companies’ accountability.

The rapid expansion of artificial intelligence (AI) technologies in 2025 has led to a substantial environmental footprint, according to new research by Dutch academic Alex de Vries-Gao, founder of Digiconomist.

His study estimates that AI-related activities released as much carbon dioxide in 2025 as the entire city of New York, highlighting the significant climate impact of this digital surge.

Notably, the research is the first to isolate AI’s environmental effects from general data center operations, revealing that AI’s water consumption now surpasses global bottled-water demand.

Specifically, AI systems could have a carbon footprint of up to 80 million tonnes of CO2 and consume approximately 765 billion liters of water, exceeding previous estimates of total data center water use by over a third.

These emissions represent more than 8% of global aviation emissions, underscoring AI’s growing contribution to greenhouse gases.

The study draws from tech companies’ own disclosures but emphasizes the need for greater transparency and accountability regarding AI’s environmental costs.

De Vries-Gao questions the fairness of society bearing these costs while technology firms reap the benefits.

Supporting this concern, advocacy groups like Foxglove highlight that the environmental burden is disproportionately shouldered by the public, as the construction of large-scale “hyperscale” data centers—some emitting as much CO2 as international airports—continues to accelerate.

The International Energy Agency (IEA) reports that AI-focused data centers consume electricity comparable to major industrial facilities and predicts that data center energy use will more than double by 2030.

The US leads in data center electricity consumption, followed by China and Europe.

In the UK, a planned data center at a former coal power station in Blyth is projected to emit over 180,000 tonnes of CO2 annually, equivalent to emissions from 24,000 homes.

Meanwhile, India’s $30 billion data center investment raises concerns about reliance on diesel generators due to grid instability, posing additional carbon risks.

Despite some tech companies like Google reporting reductions in data center emissions through clean energy adoption, challenges remain in scaling carbon-free technologies.

Google’s recent disclosures, for example, omit water usage in electricity generation for its Gemini AI, illustrating gaps in environmental reporting.

The findings, published in the journal Patterns, call for stricter regulations and greater corporate responsibility to address AI’s environmental impact as the technology continues to grow globally.