In the race to dominate artificial intelligence, tech giants are building massive data centers that guzzle electricity and, increasingly, water. But is the so-called AI water crisis as dire as headlines suggest? A closer look reveals a narrative rife with exaggeration, where evaporative cooling in data centers is painted as an environmental catastrophe, yet the numbers tell a more nuanced story. Andy Masley, in his Substack newsletter, argues that the ‘AI water issue is fake,’ pointing out that much of the water used evaporates and returns to the water cycle, not lost forever.

Drawing from recent reports, the scale of AI’s water footprint is indeed growing. A study from Nature Sustainability, published last week, projects that U.S. AI servers could consume up to 1.1 billion cubic meters of water annually by 2030, equivalent to the usage of 10 million Americans. Yet Masley counters that this evaporation isn’t ‘consumption’ in the destructive sense, as it contributes to rainfall elsewhere, unlike agricultural or industrial uses that deplete aquifers.

The Cooling Conundrum

Data centers rely on water for cooling to prevent servers from overheating, especially as AI models demand more computational power. According to an NPR report from May 2025, generative AI has driven annual increases in water use by tech firms like Google and Microsoft, with projections that U.S. data centers could use 12% of the nation’s electricity by 2028. Masley highlights that in regions like the Midwest, where Microsoft operates, water is abundant, and evaporation rates are low compared to global totals.

Critics, however, warn of localized impacts. The United Nations Environment Programme (UNEP) noted in a November 2025 article that data centers produce toxic e-waste and consume electricity often from fossil fuels, exacerbating climate issues. Masley pushes back, citing that AI’s total water use is a drop in the bucket—Google’s 5.6 billion gallons in 2022 pales against the 322 billion gallons used daily in the U.S., per USGS data.

Beyond the Hype: Real Metrics

Posts on X (formerly Twitter) amplify the debate, with users like Laura Miers warning that AI could consume up to 1.7 trillion gallons worldwide by 2027, fueling fears of resource scarcity. Yet Masley dissects these claims, noting they often conflate withdrawal with consumption; most water is returned to sources after use, not permanently removed.

A Forbes article from February 2024 echoes concerns, stating that generative AI has spiked water usage, but Masley argues this ignores efficiency gains. For instance, MIT News in January 2025 explained that while training models like GPT-3 uses significant water—estimated at 700,000 liters per model—the per-query impact is minimal, around 1/15th of a teaspoon, as Sam Altman of OpenAI has claimed.

Regional Strains and Global Perspectives

In drought-prone areas, the issue intensifies. A Food & Water Watch report from April 2025 predicts AI could use electricity equivalent to 28 million households by 2028, with corresponding water demands straining supplies in places like Arizona. Masley points to Microsoft’s data centers in Noord-Holland, Netherlands, which used billions of liters but in a water-rich region, as per a UN science-policy brief.

International efforts are ramping up. The EU’s upcoming AI Act, mentioned in a University of Illinois article from October 2024, requires reporting on environmental impacts, including water. Masley views this as positive but cautions against alarmism, noting that AI enables water-saving innovations, like smart irrigation systems that reduce agricultural waste.

Mitigation Strategies in Play

Tech companies are responding. Google and Microsoft have pledged to replenish more water than they consume by 2030, investing in watershed projects. A Cornell University study from November 2025, published in Nature Sustainability, outlines strategies to curb AI’s footprint, such as locating data centers in cooler climates to reduce evaporation needs.

Masley emphasizes closed-loop cooling systems that recycle water, minimizing net consumption. X posts from users like Vishan highlight AI’s energy thirst—every ChatGPT query uses 2.9 Wh—but Masley argues water is a secondary concern compared to carbon emissions, with AI potentially offsetting more through efficiency in other sectors.

Policy Push and Industry Accountability

Lawmakers are taking note. The Artificial Intelligence Environmental Impacts Act of 2024, introduced by Senator Edward Markey, mandates standards for assessing AI’s eco-footprint, as per the University of Illinois report. This could force transparency on water use, currently underreported in AI model cards.

Yet Masley warns that overregulation might stifle innovation. A Devdiscourse article from six days ago equates AI emissions to millions of cars, but he counters with data showing AI’s growth could be managed sustainably, citing Morgan Stanley’s projection of 106.8 billion liters annually by 2028—a fraction of global use.

Innovation as the Antidote

Emerging tech offers hope. A Newswise article from five days ago discusses AI-driven water treatment transformations, potentially revolutionizing conservation. Masley aligns this with his thesis: AI isn’t the villain but a tool for environmental solutions, like optimizing power grids to reduce overall waste.

Industry insiders see a path forward. Environment+Energy Leader reported on November 15, 2025, that AI servers could demand 245 TWh of power by 2030, but coordinated mitigation—like renewable energy integration—could achieve net-zero. Masley concludes that framing AI as a water hog distracts from real culprits like agriculture, which uses 70% of global freshwater.

The Bigger Picture: Balancing Progress and Planet

As AI evolves, the water debate underscores broader sustainability challenges in tech. UNEP’s recent story stresses the need for global standards, while X sentiment reflects public anxiety, with posts warning of surveillance and emissions tied to AI.

Ultimately, Masley’s debunking invites a recalibration: acknowledge the costs, but weigh them against benefits. With proactive measures, the AI water ‘crisis’ may prove more myth than reality, allowing innovation to flow without drying up resources.