Dan Dreyfus: The Next AI Bottleneck is Copper

Dan Dreyfus opens by framing the current moment as a historic inflection point in US economic growth. For roughly two decades, the US created enormous market value through capital-light businesses like Google, Meta, Apple, and SaaS platforms, while simultaneously offshoring critical infrastructure and manufacturing to China. Geopolitical shocks like COVID, the Russia-Ukraine conflict, and trade tensions exposed the fragility of these supply chains, causing inflation spikes that never fully reversed.

Dreyfus identifies multiple simultaneous capital cycles now underway: aerospace (Boeing/Airbus have $1 trillion in backlogs), grid modernization (parts of the US grid are over 106 years old and barely functional), data centers (now a $1 trillion per year infrastructure category), semiconductor fabs (potentially trillions), and defense spending globally. He emphasizes that none of these objectives can be achieved without critical minerals.

On China's strategic leverage, Dreyfus describes how China's April announcement cutting off exports of rare earth elements like samarium, gadolinium, and terbium brought Ford Motor Company within days of shutting down its entire production line. In response, the US government is now actively deploying equity checks, permits, and take-or-pay offtake agreements to fast-track domestic mining projects — a historically unprecedented intervention Dreyfus calls a 'vuja day' moment.

The centerpiece of his argument is copper. He lays out a stark supply-demand picture: copper demand is 30 million tons per year, growing with GDP alone would require 700 million tons over the next 18 years — equal to all copper mined in the last 10,000 years — yet only a handful of tier-one mines are coming online before the end of the decade, and it takes 7-12 years to build a new copper mine. On top of baseline demand, AI data centers (50,000 tons of copper per gigawatt, with 15 gigawatts being built per year), electric vehicles (5-6x the copper of combustion engines), renewable energy (solar uses 5x and wind 7x more copper than gas turbines), and military munitions all compound the deficit. Annual copper supply grew by only 500,000 tons last year while data centers alone will need 750,000 tons annually.

Dreyfus also highlights silver as a near-term crisis, noting a 200 million ounce annual deficit against only 600 million ounces of above-ground inventory — roughly three years before stockout — with photovoltaic solar cells as a major demand driver.

On the grid, he argues the US has not meaningfully invested in grid modernization since World War II, and that simply electrifying homes and businesses (heat pumps, EVs) — before accounting for AI — will cause blackouts, brownouts, and rising electricity prices. He notes that craft labor is arguably the single biggest bottleneck across all these build-outs, and that ironically, the blue-collar workers displaced by offshoring in the 2000s are now the most in-demand, with entry-level salaries starting at $150,000.

Finally, Dreyfus situates commodities as a hedge against dollar debasement, pointing to $40 trillion in federal debt growing at $2.5 trillion per year plus $100 trillion in unfunded social liabilities, arguing that hard assets and commodities dramatically outperformed in the analogous 1970s environment when the dollar lost 70% of its purchasing power. He concludes that copper commodity cycles typically last 15 years with hundreds of percent upside, and that this cycle is only a few years old.