The Physical Price of AI: A Looming Crisis in Power and Copper

WASHINGTON, D.C. – June 16, 2026 – The artificial intelligence boom, long celebrated for its ethereal, code-based innovations, is beginning to cast a very large and very physical shadow. While tech headlines focus on the next large language model, a more fundamental story is unfolding in power substations and copper mines. According to a new analysis by financial researcher and former government advisor Jim Rickards, the exponential growth of AI is putting an unprecedented strain on America’s power grid, a strain that runs straight through to the global supply of the raw materials required to sustain it. The story of AI’s future, it seems, may be written not in Python, but in copper.

The AI Energy Shock

The connection between your household electric bill and the AI industry is no longer theoretical. The vast server farms that power AI are voracious consumers of electricity, and their demand is skyrocketing. Peer-reviewed research from a consortium of institutions, including North Carolina State University, projects that this demand could raise U.S. power costs by a national average of 6% to 29% by 2030. For some regions, particularly data center hubs like Northern Virginia, the hike could be as high as 57%.

These are not outlier predictions. A 2024 report from the prestigious Lawrence Berkeley National Laboratory (LBNL), published by the U.S. Department of Energy, provides a stark trajectory. Data centers, which consumed roughly 4.4% of total U.S. electricity in 2023, are on track to demand between 6.7% and 12.0% by 2028. The driver is clear: AI-related servers, whose energy consumption exploded from a mere 2 TWh in 2017 to 40 TWh in 2023.

Utility providers are sounding a similar alarm. The Edison Electric Institute (EEI), an association representing all U.S. investor-owned electric companies, noted a 3% jump in electricity generation in 2024—the largest in five years—attributing it directly to the demands of AI and data centers. In response, EEI member companies are planning to invest over $1.1 trillion in grid modernization and expansion over the next five years. In Virginia, utility giant Dominion Energy projects its data center power demand will surge nearly fivefold, from 3.6 GW today to 17.3 GW by 2045. This isn't a distant problem; it's a present-day industrial mobilization, and as one Harvard Law analysis warns, the infrastructure costs risk being socialized onto consumers through higher rates.

The Copper Choke Point

Meeting this tidal wave of new electricity demand requires a massive physical build-out. You cannot simply will a more powerful grid into existence; you have to build it with steel, concrete, and, most critically, enormous quantities of copper. This is where the digital ambitions of AI collide with the hard realities of the physical world.

The global copper market is already flashing warning signs. Morgan Stanley forecasts a refined copper deficit of approximately 600,000 tonnes in 2026, the largest supply-demand gap in over two decades. This deficit is the result of a perfect storm: disruptions at major mines and a lack of new supply are colliding with surging demand not only from data centers but from the entire green energy transition, which relies on copper for everything from wind turbines to electric vehicles. S&P Global reinforces this view, projecting that overall copper demand could rise by a staggering 50% by 2030.

This reframes the central question for anyone tracking the AI sector. The most significant constraint on the AI build-out may not be the availability of capital or the supply of advanced microchips, but the physical availability of a 10,000-year-old metal. As Rickards argues, the focus shifts from asking “which AI company will win?” to the more fundamental question: “what does the entire build-out physically require, and who can supply it?” The answer increasingly points to a market where demand is programmatically certain, but supply is physically constrained.

The Regulatory Fault Line

This physical constraint is compounded by a political one. Rickards draws a provocative parallel to a market anomaly he identified during the Trump administration involving Fannie Mae. The mortgage giant’s shares were trading for pennies on the dollar after its 2008 government bailout, held down not by poor fundamentals but by a regulatory decision. When talk of privatization emerged, the stock’s value was unlocked, soaring over 1,000%. The lesson, he suggests, is that markets heavily misprice real assets when their value is artificially suppressed by government policy.

He contends that a similar dynamic is at play today with a key to the copper shortage: a massive, undeveloped U.S. copper deposit. The asset is real, the demand is undeniable, and the only thing holding it back is a regulatory roadblock he expects to shift. While the specific project is the subject of his presentation, the argument taps into a well-documented tension across the American resource landscape. Major mining projects in the U.S. often face years, if not decades, of permitting delays and legal challenges from a web of agencies like the EPA and Bureau of Land Management.

This creates a profound strategic contradiction. The very policies aimed at environmental protection and sustainability are, in this context, hindering the acquisition of materials essential for the green energy transition and the next wave of technological progress. The race for AI dominance is therefore inextricably linked to national resource policy and the political will to navigate the complex trade-offs between environmental concerns and industrial necessity. How this tension is resolved will have enormous implications for market competition and America’s ability to build the infrastructure of the future.

For strategists and investors, the implications are clear. The AI value chain extends far beyond Silicon Valley, deep into the territories of utility providers, infrastructure engineers, and mining conglomerates. The quiet, often-overlooked mechanics of power generation and commodity supply are becoming the dominant factors in a story everyone thought was about software. Understanding the ripples from these physical constraints is no longer optional; it is fundamental to grasping the future of modern advancement.