AI Is Mining More Than Data
by Hadley Chance and Mia Montoya Hammersley
Artificial Intelligence (AI) has crept into the everyday lives of tech users. Google attached “generative AI” to its search function in mid-2024; Siri has been embedded in the iPhone since 2011; ChatGPT, popular in education and business settings, was released to the public in 2022. As this tool becomes more commonplace, we need to change our perspective of “convenience.” AI does not exist in the ether, but as a physical machine that requires space, components, and resources, leading to the degradation of human rights and the planet. As the field of AI grows, its impacts can be traced backwards in the supply chain, starting with mining operations for critical minerals required to build the hardware. The run for critical materials in the U.S. and abroad is going to have harsh impacts on the environment and grim implications for environmental justice.
AI is hardware merging with software. To simplify and understand the process, an overview of hardware is necessary. In short, AI needs semiconductors, which are made out of substances that have a specific level of conductivity to support AI’s functions, with the highest mineral demand on copper. If AI continues to grow as a tech industry, the demand for copper will increase by one million metric tons by 2030 and by 3.4 million tons a year by 2050.
The Department of Energy has classified copper as a critical material. Critical materials are materials that have been designated as necessary to the “functioning of modern society.” Materials are added to the list because of potential risk to their supply chain. As AI use increases, the supply is likely to dwindle, enforcing copper’s status as a critical material. Some critics estimate that copper demand will double with the AI boom, where there is already a global copper shortage.
Along with the listing of copper as a critical material, the Biden Administration made a concerted effort to secure a foothold for the United States in the critical materials and minerals industry. The administration’s goal, met by investing more than $17 million in domestic mining projects, is to make the U.S. a major supplier of these minerals. The U.S. is the fifth largest producer of copper in the world, following Chile, Peru, China, and the Democratic Republic of Congo. At fifth place, the U.S. accounts for 6% of the world’s copper supply. Trying to bolster domestic growth, we are likely to see more and more proposed copper mines nation-wide.
This should give us pause. Copper mining in the U.S. frequently takes the form of open pits, where copper is extracted close to the surface. This method can use chemicals to leech the desired minerals out of ore or, alternatively, explosive materials and drilling are used to reach the minerals. Open pits often drill below the water table, which can lead to water contamination. This leeching process, where the chemicals are injected into ore and left to seep into the material, leads to a slurry of dissolved copper and radionucleotides, including dangerous compounds like sulfuric acid. One of the byproducts of this process is concentrated radioactive materials, known as Technologically Enhanced Naturally Occurring Radioactive Materials (TENORM). Populations who live near sites storing and producing TENORM can experience direct gamma radiation, inhalation of contaminated dust, inhalation of downwind radon, ingestion of contaminated well water, ingestion of food contaminated by well water, and ingestion of food contaminated by dust deposition. As of August 2024, the Environmental Protection Agency (EPA) is investigating regulating TENORM, but no regulations have been made.
The byproducts of copper mining have entered and severely impacted water ways, farmland, wildlife, and community health. According to one report, 92% of water treatment systems failed, leading to ground water contamination from acid and metals.
When the demand for copper skyrockets, we see more proposed mines. One example is the proposed Resolution Copper mine in Superior, Arizona. Arizona produces 65% of the country’s copper and is home to ten (soon to be twelve) copper mines. The land Resolution Copper is planning to drill, called Oak Flat, or Chi’chil Biłdagoteel in Western Apache, was held as a national forest within the Tonto National Forest. It was conveyed to Resolution Copper as a land swap under the National Defense Authorization Act of 2015.
Before it was a national forest, and for time immemorial, Oak Flat has been a sacred place for several Tribes, including the San Carlos Apache. The land has been used for religious rites such as ceremonies, to gather medicine, and to pray. Resolution Copper is currently awaiting their Record of Decision per the National Environmental Policy Act (NEPA). Copper production has not yet begun. An Indigenous organization, Apache Stronghold, filed a lawsuit against the government in an attempt to reverse the land swap. This case was filed both on the grounds of violations of the Religious Freedom Restoration Act and of an 1852 Treaty, where Oak Flat was set aside for the Apache. The potential destruction of Oak Flat provides one example of the continued consequences of mining on Tribes’ ancestral territories. In addition to treaty violations and destruction of cultural resources, Tribal consultation is neglected and desecrated.
Resolution Copper, too, is impacting water supply in Arizona. The mine is pumping more than 600 gallons of groundwater per minute to reach the copper deposits far below the water table. The mine is about an hour by car from Phoenix, where home construction was halted due to ground water supply concerns. Like many mines supported by the Biden critical minerals and materials policy, Resolution Copper stands to over-consume water in arid areas of the country. For a similar example, but with Lithium, check out Rhyolite Ridge in Nevada.
Copper mining is but one facet of environmental impacts of AI. There are several more considerations. AI, once built and functioning, is housed in large remote servers. As of 2023, these servers demand about 1-1.5% of the world’s electricity. Electricity consumption from these servers may surpass electricity consumed by some small countries. The function of electrical grids where these servers are housed, like Northern Virginia, will be put at risk, and so will their infrastructure. To put into perspective how much more energy intensive AI is, one Google search (without Gemini’s interference) will use one-tenth the electricity as one with an AI generated response. Areas supporting AI hubs are beginning to feel the increased energy demand. AI infrastructure is projected to consume more water than Denmark. AI uses water in two ways—onsite server cooling and offsite electricity generation. The water consumed does not include the water necessary for AI physical production.
Additionally, AI has a mounting waste footprint. By the year 2030, AI is projected to produce an extra 5,000 tons of E-waste. E-waste is waste from electrical equipment and have hazardous and toxic elements which negatively affect people and the environment. For AI, E-waste comes from the data servers (GPUs, CPUs, memory storage devices) using energy. Also, the servers become outdated rapidly because of the speed at which the AI industry is developing.
Most of this ends up in landfills in poorer countries across the globe. Low-income workers typically handle these hazardous and toxic E-waste materials without any protective gear. The Basel Convention, an international treaty dedicated to controlling hazardous waste disposal, mandates that developed nations no longer dump hazardous waste, including E-waste, on poorer countries. The United States has not ratified the treaty—the only developed nation not to do so.
As we move into a new Presidential administration, it is likely that Trump will reaffirm commitments to critical materials. Domestic mining will likely grow to satiate the demands of AI and other tech. According to some studies, six large copper mines will need to open a year to meet growing copper demands. Some positives as AI grows—there have been movements both nationally and internationally to moderate the environmental impacts of AI. In the U.S., a bill has been introduced to require the EPA to study environmental impacts of AI. Abroad, UNESCO has created its Recommendations about the Ethics of Artificial Intelligence, which include provisions about environmental impacts of AI. We don’t know where this winding road of AI is going, but we need to be aware of what consequences will arise from rapid, unchecked development.
Author Bios
Hadley Chance (they/them) is a 3L at Vermont Law and Graduate School. They are a senior staff editor on Vermont Law Review. They grew up outside of Philadelphia, but went to the College of Charleston, where they began to be interested in Environmental Justice. They were the co-chair of the Environmental Justice Law Society their 2L year and have been on the E-Board of Alliance, the LGBTQ+ law student group, for the past two years. They also spent the last year with Professor Montoya Hammersley in the Environmental Justice Clinic. Hadley hopes to continue this advocacy work after graduation. When not in school, they enjoy the outdoors and playing tabletop games.
Mia Montoya Hammersley is the Director of the Environmental Justice Clinic and an Assistant Professor of Law. She is a member of the Piro-Manso-Tiwa Indian Tribe, Pueblo of San Juan de Guadalupe, and a Yoeme (Yaqui) descendant. In her work, Mia has represented conservation organizations in protecting land from extractive industries, Tribes in defending and asserting their land and water rights, and communities experiencing disproportionate environmental health harms. Her chapter, “The Water-Energy Nexus and Environmental Justice: the Missing Link Between Water Rights and Energy Production on Tribal Lands” was published in the UA Press Series, Indigenous Environmental Justice, in 2020. In 2021, she was a recipient of the Young, Gifted, and Green 40 Under 40 Award by Black Millennials for Flint for her work in the field of environmental justice.