Artificial intelligence has become the defining technology of the 21st century, but its future depends on something far older and more fundamental: the soil beneath our feet. From the rare earth elements that power advanced semiconductors to the land required for massive data centers, the physical infrastructure of AI is inextricably linked to geography, geology, and national resources. As governments worldwide push for greater AI sovereignty, they are confronting a sobering reality: you cannot achieve digital independence without securing the analogue foundations first.
The resource chain of artificial intelligence
Every AI system, from a simple chatbot to a transformer model with trillions of parameters, relies on a complex global supply chain of materials and components. At the bottom of this chain are raw materials: lithium for batteries, cobalt for magnets, neodymium for hard drives, silicon for wafers, and a cocktail of rare earth elements for capacitors and lasers. The vast majority of these materials come from a handful of countries, often with fragile geopolitical ties. China, for example, controls over 60% of rare earth mining and nearly 90% of rare earth processing, giving it disproportionate influence over the production of advanced electronics, including the GPUs and accelerators used for AI training.
Meanwhile, Taiwan produces over 90% of the world's most advanced semiconductors, made at TSMC. Any disruption to that island's operations would cripple AI development globally. The U.S. and Europe have responded with initiatives like the CHIPS Act and the European Chips Act, which aim to reshore chip manufacturing. But building a fab is not enough; you also need the raw materials, the skilled labor, the clean water, the reliable electricity, and the political will to sustain decades of investment. This is what it means to begin with the soil.
From mines to models: The geography of computation
The phrase "soil beneath our feet" is not merely metaphorical. To run large-scale AI workloads, companies build hyperscale data centers that require vast tracts of land, preferably in regions with low energy costs, cool climates, and access to renewable power. Northern Virginia, for example, has become the world's largest data center market because of its proximity to fiber optic backbones, cheap hydroelectric power, and amenable zoning laws. But as demand grows, data centers are moving toward colder regions like Scandinavia and Iceland, where geothermal and hydroelectric resources are abundant.
Yet land is only one piece. The construction of a single data center consumes enormous amounts of concrete, steel, copper, fiber optic cable, and water. Each component must be sourced, transported, and assembled, often across international borders. The carbon footprint of this construction is enormous, but more importantly, it exposes vulnerabilities. If a nation cannot produce its own steel or refine its own copper, it remains dependent on foreign suppliers. This is why AI sovereignty advocates argue for a holistic industrial policy that encompasses mining, refining, manufacturing, and assembly within a country's borders.
Historical context: Lessons from the chip shortages
The fragility of the AI supply chain was exposed during the COVID-19 pandemic, when a sudden surge in demand for electronics collided with logistical disruptions. Automakers couldn't get chips; consumers couldn't buy GPUs. The price of rare earth elements spiked, and countries realized that their digital economies were built on just‑in‑time supply lines that could snap overnight. Since then, governments have poured hundreds of billions of dollars into domestic semiconductor facilities. The U.S. CHIPS and Science Act allocated $52 billion for semiconductor manufacturing and research, while the EU Chips Act aims to mobilize €43 billion in public and private investment. South Korea, Japan, and India have launched similar programs.
But semiconductor fabrication is only part of the story. The materials used in chips are often sourced from conflict zones or politically unstable regions. Cobalt from the Democratic Republic of Congo, tin from Indonesia, tungsten from China—each comes with ethical and security risks. The push for vertical integration has led some companies to explore alternative materials, such as graphene or silicon carbide, but these are years away from mass adoption. For now, the path to AI sovereignty runs through the earth itself.
Europe's strategy: The raw materials act
The European Union has been particularly aggressive in addressing the resource gap. In 2023, it introduced the Critical Raw Materials Act, which sets a target that at least 10% of the EU's annual consumption of strategic raw materials should be extracted within the bloc by 2030, along with 40% processed domestically and 25% recycled. The act identifies a list of 34 critical materials, including lithium, cobalt, copper, and rare earths. The goal is to reduce dependence on China and Russia while fostering a circular economy for electronics.
However, mining within Europe faces environmental and regulatory hurdles. Opening new mines in Scandinavia, Portugal, or Germany often triggers protests from environmental groups and local communities. The trade‑off between green energy and green AI is a delicate one. Proponents argue that without domestic mining, Europe will remain a technological colony, importing both raw materials and finished products. The soil beneath European feet contains significant deposits of lithium in Portugal and Spain, rare earths in Sweden and Finland, and cobalt in Finland—but extracting them responsibly requires new technologies and social licenses.
United States: The rare earth race
In the United States, the only functioning rare earth mine is Mountain Pass in California, owned by MP Materials. But the ore still goes to China for processing. To break this dependence, the Department of Defense has awarded contracts to companies like MP Materials and Lynas to build processing facilities on American soil. Similarly, the Biden administration has invoked the Defense Production Act to boost domestic mining and processing of critical minerals. The Inflation Reduction Act includes tax credits for battery manufacturing and mineral processing, further incentivizing domestic supply chains.
But the United States also faces a shortage of skilled workers in mining and semiconductor manufacturing. Decades of offshoring have eroded the talent base. Universities are now expanding programs in mining engineering, materials science, and chip design, but it will take years to rebuild. AI sovereignty, therefore, is not just a matter of digging—it's about educating a new generation of engineers who understand the lifecycle of technology from extraction to deployment.
India and the Global South: Leapfrogging or falling behind?
For emerging economies, the AI sovereignty debate takes on a different dimension. India, for example, has abundant rare earth deposits but lacks processing infrastructure. It also has a booming IT sector and a growing appetite for AI. The Indian government's National Mission on Interdisciplinary Cyber‑Physical Systems includes investment in semiconductor manufacturing, with a $10 billion incentive scheme. However, India remains heavily dependent on imports for both chips and raw materials.
Other nations in Africa and South America have mineral wealth but lack the political stability or capital to develop it. Countries like Zambia (copper), Chile (lithium), and the Democratic Republic of Congo (cobalt) are critical to the AI supply chain, yet they often export raw materials and get little value addition. The concept of "soil sovereignty" implies that these nations should have the right to process their own minerals and build their own AI infrastructure, rather than serving as mere quarries for the developed world. This raises complex questions about resource nationalism, trade agreements, and international cooperation.
The role of recycling and circular AI
An often‑overlooked aspect of AI sovereignty is recycling. Electronic waste contains significant quantities of rare earths, precious metals, and other materials. Currently, only about one percent of rare earth elements are recycled globally. Improving recycling rates could reduce the need for new mining and strengthen domestic supply chains. The European Union's Circular Economy Action Plan aims to increase recycling of critical raw materials, and Japan has long been a leader in urban mining—extracting metals from discarded electronics.
Several startups are working on recycling rare earths from hard drives, speakers, and electric vehicle motors. If scaled, these technologies could allow countries to become less dependent on imports. AI systems themselves could be used to optimize recycling processes, creating a virtuous cycle. However, recycling alone cannot meet the soaring demand for AI infrastructure; it must be part of a broader strategy that includes sustainable mining, efficient manufacturing, and thoughtful design.
Policy implications and the path ahead
AI sovereignty is not a binary condition—it is a spectrum. No country can be fully self‑sufficient in every material and component. The goal is to reduce critical dependencies and build resilient supply chains that can withstand geopolitical shocks. This requires international coordination on trade, environmental standards, and technology transfer. Calls for decoupling from China must be balanced with the reality that global supply chains have brought efficiency and innovation.
The next decade will see intense competition for the resources underpinning AI. Countries that invest early in mining, processing, and manufacturing will have a strategic advantage. But they must also address the environmental and social costs of extraction. The soil beneath our feet is finite, and how we treat it will determine not only the future of AI but also the health of our planet.
Ultimately, the phrase "AI sovereignty begins with the soil beneath our feet" serves as a reminder that even the most advanced digital technologies are rooted in the physical world. Leaders who ignore this reality risk building their AI ambitions on sand rather than rock.
Source: UKTN News