The Geopolitics of AI in the Americas Series Volume 5: Natural Resources and Energy Demand

Last week we examined the supply chains and infrastructure that make AI possible. But infrastructure doesn't materialize from nothing. It requires minerals pulled from the earth, energy generated at unprecedented scale, and water drawn from increasingly stressed watersheds. The physical demands of artificial intelligence are staggering, and they're about to get dramatically worse.

Every AI chip contains rare earth elements. Every data center consumes electricity equivalent to a small city. Every cooling system demands millions of gallons of water. The technology we discuss in abstract terms of compute and algorithms is, at its foundation, a voracious consumer of natural resources. And the nations that control those resources, or lack them, will find their strategic positions shaped by geology as much as by innovation.

For the Americas, this resource dimension of AI creates both extraordinary opportunity and significant risk. The hemisphere sits atop some of the world's most critical mineral deposits. But accessing and processing those resources requires infrastructure, capital, and political stability that not every nation possesses. The question isn't just where the resources are. It's who gets to develop them, on whose terms, and for whose benefit.

The Critical Minerals Foundation

Artificial intelligence runs on rare earth metals and critical minerals. These aren't exotic curiosities confined to laboratory shelves. They're embedded in everything from semiconductors to data center infrastructure, defense systems to consumer electronics. Without them, the AI industry simply cannot function.

Despite their name, rare earth elements aren't particularly rare. They can be found in geological formations around the world. What's rare is the concentration and processing capacity needed to extract them economically. China dominates rare earth production, controlling approximately 69% of global mining output and around 90% of the world's refined supply. This isn't accidental. It's the result of decades of deliberate industrial policy.

The story of how China came to dominate rare earths is a cautionary tale about strategic myopia. China's foray into rare earths meaningfully began during the 1980s, when then-leader Deng Xiaoping remarked that while "the Middle East has oil, China has rare earths." The United States once led rare earth production, with Mountain Pass in California operating as the world's top producer through the mid-1990s. But environmental pushback and regulatory pressure led to the mine's closure, and American output fell to near zero for much of the 2000s and early 2010s. China filled the vacuum, undercutting American producers through state support, lower environmental standards, and cheaper labor.

The consequences are now playing out in real time. For 19 out of 20 important strategic minerals, China is the leading refiner, with an average market share of 70%. When China imposed export restrictions on rare earths in 2010, it triggered an immediate global crisis and accelerated efforts by the United States, European Union, and Japan to diversify supply chains away from Chinese control. Those efforts have produced some progress, but the concentration remains formidable.

For the Americas, the challenge is clear: the hemisphere needs critical minerals to build AI infrastructure, but it currently depends on China for the refined materials that turn raw ore into usable components. Breaking this dependency requires not just mining but processing, refining, and magnet manufacturing capabilities that don't yet exist at meaningful scale outside Asia.

Copper: The Unsung Critical Mineral

While rare earths receive most of the attention in discussions of AI supply chains, copper may ultimately prove equally important. Data centers require massive amounts of copper wiring for power distribution and signal transmission. Chile is providing essential materials for AI data centers and computing infrastructure, a rapidly growing source of copper demand.

Chile retained its position as the world's largest copper producer in 2024, producing 5.5 million tons. That's roughly a quarter of global output, making Chile indispensable to any nation building AI infrastructure at scale. The country's copper reserves are the largest globally, concentrated in the Atacama Desert where operations like Escondida, the world's largest copper mine, continue to underpin Chile's export economy. But Chile's dominance is being challenged. The Democratic Republic of the Congo secured 14% of the global copper market in 2024, while Peru accounted for 12%. Peru ranks among the world's top copper producers, with output expected to grow as foreign investment, predominantly from Chinese companies, expands operations. This introduces a geopolitical wrinkle: as the Americas seek to build AI supply chains less dependent on China, key mineral inputs are increasingly flowing through Chinese-controlled mining operations even within the hemisphere.

The Lithium Triangle

South America's lithium deposits represent another critical resource for AI development, though lithium's primary role is in battery technology rather than AI chips directly. The strategic importance is indirect but real: energy storage systems that support renewable power generation for data centers depend on lithium-ion batteries, and the nations that control lithium supply will have leverage over the energy transition that AI demands.

Argentina, Bolivia, and Chile collectively hold what's known as the Lithium Triangle, a region collectively holding 58% of the world's identified lithium resources. Argentina is the most dynamic player in this space. Argentina ranks as the fifth-largest lithium producer globally, with 2024 output reaching 74,600 tonnes of lithium carbonate equivalent, a 62% year-over-year increase. The country expects to boost production by 75% in 2025 and, if current investment commitments are fulfilled, could potentially become the world's second-largest lithium producer within a decade.

But Argentina carries geopolitical risk. Erratic government policies and shifting regulatory frameworks have historically deterred sustained foreign investment. The current administration under President Milei has introduced investment incentive regimes designed to attract capital, but consistency over time remains the question. Bolivia possesses enormous lithium reserves but lacks the means of production, and resource nationalism has complicated efforts to develop them, creating a behind-the-scenes power struggle among state actors seeking to deepen ties with Bolivian resources.

Chile produces lithium alongside its copper dominance, with output expected to reach approximately 305,000 tonnes of lithium carbonate equivalent in 2025. Chile's approach treats lithium as a strategic national resource, restricting private ownership of mining concessions in ways that contrast sharply with Argentina's more open framework.

Water: The Hidden Constraint

We discussed water briefly in Volume 4 when examining data center siting, but the resource dimension deserves deeper examination. Water isn't just a cooling medium for data centers. It's a finite strategic asset that AI infrastructure competes for alongside agriculture, municipalities, and ecosystems.

High-density AI racks generate heat that requires enormous volumes of water for thermal management. As data centers scale to support frontier AI training runs, their water consumption becomes a serious constraint in arid regions. This creates a geographic paradox: many of the locations most attractive for data centers, due to cheap land, low taxes, and available power, sit in water-stressed areas.

The Americas have water-rich regions that could attract AI infrastructure. Canada possesses abundant freshwater resources. The Pacific Northwest of the United States combines water availability with cool climates ideal for data center efficiency. Certain parts of Brazil, particularly in the northeast where hydroelectric capacity is strong, offer similar advantages. But water politics are local in ways that energy politics often aren't. Communities adjacent to proposed data centers increasingly push back against facilities that compete for water essential to drinking supplies and agriculture. This "Social License to Operate" challenge means that even regions with abundant water may face political resistance to AI infrastructure buildout if communities perceive the technology as threatening their access to this most fundamental resource.

Energy: The Ultimate Constraint

Everything else is secondary to energy. You cannot run data centers without massive amounts of electricity. You cannot train frontier AI models without power consumption that rivals small municipalities. The energy dimension of AI's resource demands dwarfs everything else we've discussed.

This isn't a future problem. It's happening now. As projects like the 10GW Stargate campuses break ground, the metrics for success have shifted from simple Power Usage Effectiveness to Water Usage Effectiveness. The scale of energy demand AI creates is reshaping the economics and geopolitics of power generation across the hemisphere.

United States

The United States currently operates as a net exporter of oil and natural gas, a position achieved through the shale revolution and regulatory relaxation. The country maintains three major power grids capable of withstanding significant energy demands, placing it in the top tier globally for energy infrastructure.

But energy abundance doesn't automatically translate to AI-ready power. Nuclear energy remains stigmatized in public discourse despite being a low-carbon solution capable of providing the baseload power AI demands. The United States hosts considerable wind and solar capacity alongside fossil fuel generation, but integrating intermittent renewable sources with the consistent, high-density power requirements of data centers remains a technical and economic challenge.

The irony is sharp: building AI infrastructure at the scale the United States envisions requires either massive expansion of nuclear capacity, continued reliance on fossil fuels, or grid management innovations that don't yet exist at necessary scale. Each option carries political and environmental costs that will shape where and how American AI infrastructure develops.

Canada

Canada's energy profile is arguably the most AI-friendly in the hemisphere. Hydroelectric power provides abundant, reliable, carbon-neutral baseload electricity, particularly in Quebec and British Columbia. This has already attracted significant data center investment, as companies seek to demonstrate environmental responsibility alongside operational efficiency.

Canada also possesses substantial nuclear capacity and natural gas operations, though the latter operate under stricter environmental regulations than their American counterparts. The combination of clean energy, cool climate, and political stability makes Canada an increasingly attractive location for AI workloads that are both energy-intensive and environmentally sensitive.

The strategic question is whether Canada can capture enough economic value from this advantage to justify the investment it requires, or whether the infrastructure buildout will primarily benefit American technology companies operating Canadian facilities.

Mexico and Latin America

Mexico maintains a mixed energy system dominated by state-owned utilities. Power constraints represent one of the most significant bottlenecks to Mexico's AI ambitions, as discussed in Volume 4. Supporting the data centers and manufacturing operations that Mexico's strategic position demands will require substantial investment in generation and distribution capacity that the current system struggles to deliver.

Brazil hosts a hydroelectric base supplemented by growing wind and solar capacity. But Brazil has experienced critical blackouts and droughts that have periodically stressed its energy system, raising questions about reliability at the scale AI demands. Chile is witnessing a significant renewables buildout, but grid infrastructure limitations and price volatility create energy risks for large-scale industrial consumers.

Paraguay presents an interesting case. The Itaipu Dam, shared with Brazil, was once the largest hydroelectric facility in the world and continues to generate enormous quantities of clean power. This has made Paraguay attractive for large-scale computing projects that require reliable, carbon-neutral electricity at competitive costs. For a small nation, hosting AI computing infrastructure powered by hydroelectric capacity represents a potential economic leapfrogging opportunity.

Venezuela: Resources Without Capacity

No discussion of Americas resources would be complete without Venezuela, which presents perhaps the starkest illustration of the gap between geological wealth and strategic capacity.

Venezuela is sitting on around 300 billion barrels' worth of "proved" oil, the largest reserves of crude oil in the world. This compares with Saudi Arabia's approximately 260-270 billion barrels. Venezuela also ranks ninth globally in natural gas reserves, with deposits totalling around 5.5 trillion cubic metres.

But these reserves have proven largely inaccessible due to a combination of factors. The oil extracted from Venezuela's reserves is largely a heavy, sulfurous or "sour" grade that doesn't flow as nicely up the well bore to reach the surface and, because of its high sulfur content, more quickly corrodes equipment. Extracting and refining this oil requires specialized expertise and infrastructure that Venezuela has systematically lost through decades of mismanagement, brain drain, and underinvestment.

Despite possessing an estimated 303 billion barrels of extra-heavy crude oil reserves, Venezuela ranked just 21st in oil production in 2024, producing about 960,000 barrels per day. At its peak in the 1970s, the country pumped 3.5 million barrels per day. The collapse represents one of history's most dramatic examples of how political instability can render geological wealth strategically irrelevant.

Venezuela is unlikely to contribute meaningfully to an AI industry in the near term. Its energy infrastructure is deteriorating, its technical workforce has emigrated, and international sanctions further constrain development. The country's trajectory may instead deepen long-term instability as resource wealth remains locked underground while economic conditions worsen.

The Hemispheric Resource Vision

The Americas collectively possess an extraordinary endowment of the resources AI requires. The question is whether the hemisphere can organize itself to capture this wealth strategically rather than simply exporting raw materials while others capture the value-added portions of the supply chain.

The United States brings a surplus of energy and some rare earth deposits, though public opposition to data center siting remains a growing concern in many communities. Canada contributes nickel, cobalt, and a reliable clean power grid coupled with exceptional AI research capabilities. Mexico possesses copper, silver, and the manufacturing base for semiconductor assembly, testing, and packaging that major American companies are already beginning to trust.

Within South America, the picture is more differentiated. Argentina holds massive lithium reserves but requires sustained political stability to attract the investment needed to develop them. Chile leads in copper production and possesses significant lithium output, positioning it as a critical mineral supplier for the hemisphere. Brazil combines a large economy capable of self-investment with growing energy production capacity. Despite its membership in BRICS and good relations with China, Brazil has expressed motivation to participate in a less China-dependent rare earth supply chain, a signal worth watching carefully.

Paraguay's hydroelectric capacity makes it suitable for large-scale computing projects. Bolivia is rich in resources but lacks the production infrastructure to develop them. Peru ranks among the world's top copper producers, with output expected to grow driven largely by Chinese investment, introducing the geopolitical complications we've discussed throughout this series.

The ideal outcome would see the Americas develop an integrated resource and energy supply chain that feeds the AI industry while capturing value at every stage, from mining through processing, manufacturing, and deployment. Raw material exports alone won't build strategic capacity. The hemisphere needs to move up the value chain, developing the refining, processing, and manufacturing capabilities that transform geological wealth into industrial power.

What Comes Next

We have now covered the economic foundations of AI as a strategic industry, its employment implications, the supply chains and infrastructure it requires, and the natural resources and energy that sustain it. These four volumes have established the material and economic dimensions of AI geopolitics in the Americas.

Next week we turn to something different. We turn to the battlefield. Military applications of AI are expanding at a pace that would have seemed fantastical a decade ago, and they are reshaping what warfare means, who can wage it effectively, and how smaller nations might compete against adversaries with far larger conventional forces. AI weaponry is an unfortunate reality of the human species. We have been at war for a long time, and the military has always been at the heartbeat of innovation. Next week, we examine what happens when that innovation becomes artificial intelligence.

This is Part 5 of a 10-part series on The Geopolitics of Artificial Intelligence in the Americas by Core Geopolitical Insights LLC. Follow along each week as we explore how this transformative technology is reshaping power, prosperity, and security across the Western Hemisphere. | Next week: AI Weaponry