Crushing and Grinding
Once extracted, raw ores are transported to processing facilities where they are first crushed and ground into fine particles. This step increases the surface area of the material and prepares it for chemical separation. While seemingly benign, grinding consumes significant amounts of electricity and water and generates dust and fine particulate pollution, especially in dry regions. Each step requires significant energy and often involves hazardous substances. For example, lithium extracted from brines or hard rock must be refined into battery-grade lithium carbonate or hydroxide, which entails high water and chemical use. Similarly, rare earth elements require multi-step refining using acid baths and ion exchange processes that can generate toxic waste if not managed properly.

Chemical Refining
Refining separates the valuable metals from other elements in the ore. This often involves solvent extraction, acid leaching, and precipitation, using substances such as sulfuric acid, hydrochloric acid, or sodium hydroxide. These processes are critical for producing battery-grade lithium, cobalt sulfate, and rare earth oxides, but they generate toxic tailings and chemical waste that are difficult to safely store or treat.

Thermal Processing
Some metals require high-temperature treatment to be purified. For example, smelting is used to extract nickel and cobalt, releasing sulfur dioxide, particulate matter, and greenhouse gases into the atmosphere. Rare earth elements often require calcination, a high-energy step that transforms raw concentrates into usable oxides.​

Water Use and Contamination
Many manufacturing steps require large volumes of water. In lithium refining, brine is pumped from underground aquifers and evaporated in open-air ponds. This has raised serious concerns in arid ecosystems like the Atacama Desert, where water use can outpace natural recharge and threaten local livelihoods.​

Component Production
Once refined, materials are processed into intermediate goods such as silicon wafers, cathode materials, and rare earth magnets. Producing a single 300mm wafer can use over 2,000 liters of ultrapure water. These components are then shipped globally to be assembled into chips, batteries, and hardware systems for AI deployment.​

The extraction of minerals foundational to artificial intelligence systems highlights a set of long-standing structural asymmetries: between those who determine the direction of technological progress and those whose environments and livelihoods are affected by its material demands. While the importance of minerals like lithium, cobalt, and rare earth elements to AI systems is widely acknowledged, the full implications of how they are sourced—ecologically, politically, and socially—often remain underexplored in mainstream AI discourse.

This stage raises important questions about who has the authority to define acceptable trade-offs in the name of innovation. Extraction is rarely a neutral activity; it reshapes landscapes, alters relationships to land, and often transforms the rhythms of life for nearby communities—especially when governance structures lack transparency or meaningful public participation. These dynamics are shaped not only by local contexts, but by broader global supply chain pressures and international market logics.

A planetary justice lens calls attention to the fact that these processes affect more than human communities. The disruption of ecosystems—through habitat destruction, water table depletion, or pollution—also impacts more-than-human life in ways that are difficult to fully measure but increasingly urgent to consider. This perspective invites us to reflect on how extraction decisions are made, and whose values and futures are centered in those decisions.

Ecological breakdown → The refining of critical minerals produces significant amounts of toxic sludge, acid effluent, and airborne pollutants. Tailings from rare earth processing, for example, can contain radioactive materials and heavy metals that contaminate soils and water. In Baotou, China, rare earth refining has produced a toxic tailings lake visible from space, and residents have reported soil infertility and elevated rates of respiratory illness. In Southeast Asia, aquatic ecosystems near nickel refineries have suffered from sedimentation, acidification, and biodiversity loss. In the Philippines and Indonesia, concerns have been raised over forest loss and contamination of coastal ecosystems linked to mineral refining and smelting facilities.

Climate breakdown → The high-temperature processes involved in smelting, roasting, and calcination are carbon-intensive. Producing battery-grade lithium carbonate, for instance, emits large amounts of CO₂ per ton refined. In the Atacama Desert in northern Chile, lithium processing is powered by fossil fuels and uses vast quantities of water, exacerbating regional water scarcity. Local communities--many of whom are Indigenous--have raised alarms about shrinking aquifers and the ecological fragility of the salt flats. At the same time, chip fabrication plants--concentrated in countries like Taiwan, South Korea, and the U.S.--require massive energy inputs, often drawn from carbon-intensive grids, contributing to the same climate challenges AI is purported to address.

Labor rights and safety → Workers in materials manufacturing facilities frequently face exposure to hazardous substances including hydrofluoric acid, sulfur dioxide, and organic solvents. In many cases, facilities lack adequate protective gear, air filtration, or labor protections. In many manufacturing facilities, especially in regions with less stringent labor regulations, workers may lack adequate protective gear and access to proper ventilation systems. This increases the risk of exposure to harmful substances and contributes to unsafe working environments. ​ Hydrofluoric acid, in particular--used in cleaning and etching silicon wafers--can cause severe injuries or death upon contact, underscoring the need for rigorous workplace safety standards.

Resource colonialism → Communities near processing hubs--such as Baotou (China), Norilsk (Russia), and regions in Indonesia, and the Philippines--often live with the environmental legacy of mineral refining: air and water pollution, radioactive waste, and land degradation. Yet the economic value derived from refining minerals into high-tech components--rare earth magnets, cathodes, and semiconductors--is largely captured by multinational corporations and technology firms based in the Minority World. The global architecture of AI development continues to reflect long-standing patterns of resource colonialism, where material burdens are externalized while value is concentrated in the hands of a few.

The capacity to refine and process critical minerals essential to AI systems--such as lithium, cobalt, and rare earth elements--is highly concentrated in a few countries, creating strategic dependencies across the global supply chain. China plays a particularly significant role, accounting for a dominant share of global lithium processing, cobalt refining, and rare earth separation, giving it outsized influence over the availability and cost of materials used in AI hardware.​

While this concentration has enabled efficiencies of scale and investment in processing infrastructure, it has also raised concerns about vulnerability to trade disruptions, export controls, and geopolitical tensions. In response, several governments are seeking to rebalance their exposure to concentrated refining capacity. The United States, for example, has introduced industrial incentives and procurement standards through the Inflation Reduction Act to encourage domestic and allied processing projects. Similarly, the European Union’s Critical Raw Materials Act proposes expanding refining capacity within Europe to reduce reliance on external suppliers.

However, these efforts raise complex trade-offs. Diversifying supply chains is not only a matter of logistical resilience or national security--it also has implications for environmental impact, labor standards, and community participation. Building new processing facilities in alternative jurisdictions may shift the location of risk without addressing the root causes of extraction-driven harm or the global inequities embedded in current supply chain models. Without strong regulatory alignment and international cooperation, reshoring or “friend-shoring” strategies risk reproducing the same extractive dynamics under a different flag.

A planetary justice lens requires us to look beyond efficiency upgrades or pollution controls and interrogate the broader systems that shape who benefits from AI’s expansion--and who bears its risks. While the environmental and labor consequences of materials manufacturing are increasingly documented, they remain peripheral to most mainstream AI governance debates. Often, discussions center on algorithmic fairness or energy consumption, leaving out the communities and workers most directly affected by the production of AI’s physical infrastructure.

The harms faced by communities in places like Baotou, Norilsk, or the Atacama are not isolated incidents--they are symptoms of a global economic model that concentrates value creation while distributing ecological and health burdens along existing lines of inequality. These cases highlight the need for more than safeguards or certifications: they call for redistribution of voice, agency, and accountability.

A planetary justice approach insists that the people living near refineries, working in chip plants, or losing access to clean water due to industrial processes must be treated as stakeholders, not externalities. It emphasizes the importance of procedural justice--ensuring meaningful participation in decisions about infrastructure development, industrial zoning, and benefit-sharing--and demands that AI governance frameworks incorporate material, spatial, and historical dimensions of injustice.

Rather than viewing AI’s material base as an unavoidable cost of innovation, planetary justice invites us to ask: how can we build technological systems that do not require sacrifice? What would it mean to develop AI hardware supply chains rooted in ecological stewardship, regional equity, and long-term care?