Just beyond the last row of Bayview’s Silver Terrace neighborhood’s brightly colored houses, a pristine building that reads, “Digital Reality” emerges. This is one of the Bay Area’s 270 data centers—part of a global network of 11,000 facilities that store and process digital information. Like many data centers nowadays, this one houses artificial intelligence (AI) data, a technological development that generates knowledge and performs tasks mimicking human intelligence.
Despite its nondescript exterior, Digital Reality’s data center—and the industry it belongs to—comes at a steep environmental cost. Data centers consume heavy amounts of energy and water, often drained from underprivileged communities, such as Bayview. Legal measures and mitigation efforts have been made to reduce AI’s environmental cost. However, as a national energy crisis looms, the question remains: Can AI’s rapid growth truly be upheld sustainably?
From 2022 to 2023 alone, power requirements for North American data centers doubled, rising from 2,688 megawatts to 5,341 megawatts—largely due to AI’s immense need for data storage and processing. AI’s energy demand is supplied by the power grid, which draws from a mix of fossil fuels, natural gas, solar and wind energy. A 2021 study by Google and UC Berkeley revealed that training a single AI model consumes 1,287 megawatt hours of electricity—enough to power 120 U.S. homes for an entire year. This demand is expected to rise further, with utility companies predicting a 20% increase in energy consumption by 2035.
This level of consumption is not just costly, it’s also environmentally unsustainable. According to a Morgan Stanley report, within the next five years, data centers could account for 40% of annual U.S. carbon emissions.
Beyond electricity, AI relies heavily on water. Data centers use vast amounts of clean drinking water to cool their servers and prevent overheating, often drawn from the same municipal supplies that serve local communities. A single data center can consume up to five million gallons of fresh water daily—the equivalent of thousands of households’ daily usage. Given that only 0.5% of Earth’s water is fresh and accessible, AI’s growing demand for water threatens to worsen existing water crises.
Data centers also generate massive amounts of electronic waste, as they rely on minerals such as aluminum, silicon and copper that are extracted through energy-intensive mining. The mining process produces toxic waste that can contaminate local water supplies, while AI hardware is often improperly discarded, polluting soil and waterways.
The burden of AI’s resource consumption falls disproportionately on lower-income communities, where tech companies frequently place data centers due to lower regulatory oversight. Data centers often exploit these neighborhoods’ resources while obscuring the extent of their water and energy usage through nondisclosure agreements, preventing communities from understanding the data center’s full impact.
Digital Reality’s presence in Bayview exemplifies this issue. Once home to the Yelamu and Ramaytush Ohlone peoples, Bayview later became a hub for Chinese immigrants, Southern Europeans and today, a predominantly Black population. Similar data centers are located in historically underserved neighborhoods, such as Oak Center in Oakland and Oak Grove in San Jose, raising ethical concerns about environmental justice and resource allocation.
Globally, much of the e-waste generated by AI is shipped to countries in the Global South, including China, Ghana and Pakistan, where weak environmental regulations and low labor costs allow corporations to dispose of hazardous materials with minimal accountability. This reinforces a cycle where the world’s most vulnerable populations disproportionately bear AI’s environmental and social costs.
Yet, for the average consumer, these issues remain largely unnoticed. “Unless you’re tapped into this sort of thing, you’re probably not paying a whole lot of attention,” Andy Uhler, energy journalism fellow at the University of Texas at Austin, said.
photo by Frances Carlson ’26
Efforts to reduce AI’s environmental impact are underway. One approach is increasing efficiency. DeepSeek, a new AI model from China, has gained attention for requiring fewer semiconductor chips, reducing traditional AI models’ use of 16,000 chips to just 2,000—lowering energy consumption while improving efficiency.
Many tech giants have also committed to replenishing the natural resources they consume. Apple, Google and Meta have pledged to restore 100%-120% of the water they use by investing in conservation projects. Meanwhile, Amazon’s Right Now Climate Fund invests in reforestation, while Apple partners with Conservation International to protect forests in South America and Africa.
Outside of the tech industry, lawmakers are working to contain data centers’ environmental impact by increasing tech companies’ need for transparency. In California, Assembly Member Rebecca Bauer-Kahan introduced AB 222, a bill advocating for data centers and utility workers to find solutions to reduce AI’s energy costs so Californians do not bear unreasonably high costs. “This bill is about ensuring transparency, accountability and fairness as we build the infrastructure for AI while protecting our environment and California consumers,” Bauer-Kahan said. Meanwhile, Massachusetts Senator Edward Markey introduced the Artificial Intelligence Environmental Impacts Act of 2024, requiring AI developers to disclose their environmental footprint.
“We need to do a better job of allowing for transparency. I think we probably think that it’s too much for the consumer to have to worry about,” Uhler said.
Ironically, AI itself may help mitigate its environmental damage. Yale’s Expeditions in Computing Program aims to reduce AI’s carbon footprint by 45% through advanced carbon modeling and AI-driven energy efficiency research. AI is also being used for disaster prevention, weather modeling and pollution tracking.
However, recent political developments may shape AI’s future in unexpected ways. In a recent executive order, President Donald Trump emphasized AI’s role in national security and competition with China. His administration launched the $500-billion Stargate Project to expand AI infrastructure, expected to consume 15–25 gigawatts of power—enough for 19 million homes—which could further drive up electricity costs.
On January 21 earlier this year, Trump also declared a national energy emergency, suspending environmental regulations and fast-tracking the approval of 600 energy projects without environmental review. “We will drill, baby, drill,” he said during his inauguration, alluding to an aggressive push for fossil fuel consumption and leaving the future of AI’s energy use uncertain.
While the future of AI remains unclear, the average consumer can take steps to mitigate its harmful impacts, starting with simply being aware.
“If you want to be a really conscious consumer, you have to kind of dig in, which a lot of us aren’t willing to do. Even if the consumer ends up not caring, at least you’re a little bit smarter about exactly what goes into the things that you’re using,” Uhler said.
While using AI for simple tasks may seem harmless, this small step toward convenience could mean a significant step toward environmental instability.
+ There are no comments
Add yours