Artificial intelligence has a dirty secret: the graphics processing units (GPUs) that power it are leaving a trail of environmental destruction. While tech companies race to cram hundreds of thousands of these chips into data centers worldwide, communities are grappling with the consequences—rising energy bills, polluted air, strained water supplies, and growing mountains of electronic waste.

Nvidia, once a niche chipmaker and now the world's most valuable company, boasts that its “gaming breakthrough” in 1999 laid the foundation for modern GPUs. But the same hardware that enables immersive video games and AI chatbots also consumes vast resources at every stage of its life. “There are massive hardware developments that start because of games,” says Catherine Flick, a professor of ethics and games technology at University of Staffordshire. “All these sorts of things that have quite significant ethical issues, a lot of these start with games.”

The GPU's journey from raw materials to data center to landfill is fraught with environmental harm. Mining for copper, tin, and other metals often leads to acid drainage that contaminates water sources. Semiconductor fabrication uses toxic chemicals—including per- and polyfluoroalkyl substances (PFAS), known as “forever chemicals”—and has left a legacy of Superfund sites in Silicon Valley. And once deployed, GPUs in data centers guzzle electricity and water, while their manufacture releases greenhouse gases and air pollutants.

Now, with AI driving an unprecedented demand for GPUs, these impacts are accelerating. In the United States, data centers are being built at a feverish pace, often in neighborhoods already burdened by pollution. The NAACP has sued xAI (now doing business as SpaceXAI) over air pollution from gas generators at its data centers. “It’s not happening in some faraway country, but in the same places where big American tech companies are courting more affluent consumers,” notes one environmental journalist.

Hitting Close to Home

While GPUs have been around for decades, AI has thrust them into the spotlight in a way that directly affects Americans. The US has more data centers than any other country, and many more are planned. As tech companies race to expand a new generation of hyperscale facilities, communities are grappling with the prospect of having these hulking warehouses as neighbors.

Some venture capitalists have blamed AI's slow consumer uptake on bad press about its environmental impact. But that argument draws skepticism. “Isn’t it nice to have the environment as a scapegoat?” Flick says. Ashley Striblet, a product strategist who works in consumer AI, adds that many people understand the environmental costs of fast fashion or consumer technology but still buy those products because they see value. “I think most people know that ordering clothes from Shein or Amazon is not the most environmentally conscious thing to do. The cost savings from buying cheap clothes turn out to be enough of a tradeoff.” With AI, however, consumers haven't yet seen the benefits that CEOs promise. “It’s just rubbish what they’re saying that their technology can do,” Flick says. “It’d be nice if we all had personalized butlers, but how do you get from point A to point B? It’s not with what we’ve got.”

Meanwhile, the costs are becoming visible. Data centers raise utility bills and create pollution, hitting low-income neighborhoods and communities of color hardest. The NAACP has warned tech companies to “be on alert” as it helps local groups mount campaigns.

Growing Globally, Impacting Locally

Shaolei Ren, an associate professor at UC Riverside, has seen the costs communities pay. Growing up in a coal-mining region of northern China, he remembers keeping windows shut to keep out black carbon and storing water in tanks to ration throughout the day. Now he studies data centers' impact on air quality and water resources—costs invisible each time a user enters a query into a chatbot.

“The energy needed to train a model like Meta’s Llama 3.1 can produce as much air pollution as driving from New York City to Los Angeles and back 10,000 times,” Ren and colleagues estimated in a preprint study. Public health costs from AI adoption could reach over $20 billion by 2028, with 1,300 premature deaths annually from air pollution by 2030. For comparison, gaming in the US uses about 34 terawatt-hours (TWh) per year, while AI servers in data centers already consume over 40 TWh and could grow to between 165 and 326 TWh by 2028.

Water is another hidden cost. AI was estimated to use between 312.5 billion and 764.6 billion liters of water in 2025—comparable to global bottled water consumption. The demand is “spikey,” Ren notes, with data centers using 6 to 10 times more water during peaks, and some massive projects needing up to 30 times more. Meeting peak water demand through 2030 could cost $10 billion, burdening underfunded community water systems.

From the Cradle

To understand GPUs' full environmental impact, researchers like Sophia Falk and David Ekchajzer have thrown tens of thousands of dollars' worth of chips into industrial blenders. Their work reveals the material footprint: an Nvidia A100 is 90% heavy metals, primarily copper. Training GPT-4 could require up to 8,800 A100 GPUs, ultimately extracting and discarding up to 7 tons of toxic elements. The GPU chip itself dominates 10 of 16 environmental impact categories, with manufacturing accounting for most public health risks from toxic chemicals and nearly all cancer risk.

Chip manufacturing has left a toxic legacy. Santa Clara County has more Superfund sites than any other US county due to chemical leaks. Some chemicals linked to miscarriages were phased out in the US in the 1990s, but manufacturing moved to Asia, where similar health risks emerged later. Now generative AI is driving a rush to build new fabs, including TSMC's Arizona facility producing Blackwell chips. The energy consumed by AI chipmaking jumped 350% from 2023 to 2024, with a fourfold rise in carbon pollution.

In Phoenix, residents like Cheryl Orosco worry about forever chemicals contaminating water sources. “It’s terrifying,” she says. “This is our forever home.”

To the Grave

Alex de Vries-Gao, who tracked crypto mining's energy use on Digiconomist, now focuses on AI's e-waste. He estimates AI servers could create between 0.131 million and 0.225 million tons of e-waste each year by 2030—comparable to a country like Denmark. Only 22% of global e-waste is collected and recycled. The rest enters the informal sector, often in lower-income countries where workers manually dismantle electronics without protection. AI servers contain valuable materials like gold, silver, and platinum, but also toxic lead and chromium. Around 58% of AI e-waste comes from North America, but the US has not ratified the Basel Convention limiting hazardous waste trade, allowing e-waste to be shipped abroad.

“Whether you’re talking about crypto miners or big tech companies, they don’t take responsibility,” de Vries-Gao says.

Gaming Out the Way Forward

Researchers stress that while data centers account for only 1.5% of global electricity, their exponential growth demands scrutiny. “If we want to electrify the transport system, it is being jeopardized by the fact that we are using additional electricity for AI rather than for transport,” Ekchajzer says. The question is whether generative AI's benefits outweigh societal costs. “Generative AI is totally one of those classic solutions looking for a problem,” Flick says.

Solutions include extending GPU lifespans, downcycling obsolete cards, and designing more efficient chips. Nvidia’s Blackwell Ultra is 50 times more efficient than its Hopper predecessor. But efficiency can backfire if it encourages even more resource use. “I have a question of sufficiency: where do we stop?” Falk asks.

Environmental groups want more accountability. Greenpeace East Asia pushes Nvidia to clean its supply chain, noting that the company avoids public scrutiny despite a $4 trillion market cap. Meanwhile, gamers—who are often more engaged with climate action—could advocate for sustainability. “Games are very powerful tools for narrative and telling stories, and shaping how people think about things,” Flick says.

Ultimately, addressing consumer concerns means “directly address and be empathetic to consumers’ concerns, so that you’re able to actually develop the solutions,” Striblet says. The path forward requires balancing AI's promise with the real-world costs that communities bear today.

Source:The Verge News