The copper squeeze: AI, electrification, and a decade of mines the world isn't building
On 10 June 2026, Dan Dreyfus, a partner at Fortnite Capital with a quarter-century in commodities, walked into the allin studio and laid out a thesis that the next industrial cycle will be defined not by software valuations but by what gets pulled out of the ground. The argument is structural: the United States spent two decades building a capital-light economy — trillions in market cap from companies like Google, Meta, and WhatsApp (acquired for $30 billion with twelve employees) on top of a hollowed-out industrial base — while "literally tearing down all of our critical infrastructure and moving it overseas to China." The country is now colliding with a demand shock it cannot meet with the supply it has.
The math, in Dreyfus's telling, is unforgiving. Global copper demand runs at roughly 30 million tons a year — about 4 million tons of that from recycled scrap and 26 million tons freshly mined. If demand merely tracks GDP growth, the world will need 700 million additional tons of copper over the next 18 years, equivalent to all the copper humanity has mined in the last 10,000 years. That implies bringing five new tier-one mega mines online every year for nearly two decades. The number of tier-one copper projects actually moving toward first production this decade, Dreyfus says, can be counted on one hand. Permitting and construction timelines run 7 to 12 years. The gap is not a forecast — it is a calendar problem already in motion.
The demand side is the easier half to map. A single gigawatt of AI data center capacity requires 50,000 tons of copper. The US is on track to build 15 gigawatts of these facilities a year, which works out to 750,000 tons of copper demand from data centers alone — against only 500,000 tons of total global copper supply growth last year. Add the reshoring of semiconductor fabs (a $750 billion capex cycle that Dreyfus expects to be "measured in the trillions"), the combined Boeing-Airbus backlog of roughly $1 trillion over the next decade, the space economy, defense procurement, and grid hardening, and the US is looking at a trillion-dollar-plus capital cycle every year for thirty years. As Dreyfus put it: "We're going to be measuring human progress by how much electricity we consume."
The supply wall
The supply side is where the optimism runs out. China announced in April 2026 that it would cut off exports of several critical minerals to the US, including samarium, gadolinium, terbium, dysprosium, lutetium, scandium, yttrium, and silver. The disruption was not theoretical. Ford Motor Company came within days of shutting its entire production line because of the samarium cobalt magnet cutoff, and McDonnell Douglas faced the same near-shutdown. These were not minor parts in minor assemblies — samarium cobalt magnets are embedded in actuators, sensors, and motor components across the auto and aerospace supply chains. The margin between continued production and a hard stop was measured in weeks of inventory.
The near-miss has, finally, produced a policy response. The US government — through what Dreyfus describes as a coordinated posture between the Department of War and the Department of Energy — is now approaching small resource owners with a three-pillar deal that he says is unprecedented in his 25 years in commodities. First, an equity check from the government to fund mine conversion and rehabilitation. Second, an expedited permit under what amounts to a national-security rationale. Third, a take-or-pay offtake agreement with a minimum floor price, designed to guarantee internal rates of return for the operator while preserving the upside if commodity prices climb. The structure acknowledges something that pure private capital cannot: the first mine out of the gate in a new district carries exploration risk that no bank will underwrite, and the second mine benefits from the first mine's data. Public capital has to absorb the pioneer loss so private capital can take the harvest.
Dreyfus is honest about the timeline. "It will take at least 10, probably 20 years to catch up to China's grip on critical minerals." China spent two decades building processing dominance, and that capacity does not replicate from a signature and a press release. Permitting reform, equity injections, and floor-price offtakes compress the front end of the curve — they do not eliminate it. The structural read is that the policy is necessary and insufficient, and the gap between US demand and US supply for the rest of this decade is going to be closed, if at all, by imports from jurisdictions that are not particularly friendly.
The other squeeze
Copper is the headline, but it is not alone. Silver is running a 200-million-ounce annual supply deficit against roughly 1.2 billion ounces of demand — a gap that draws down the above-ground inventory, now sitting at around 600 million ounces, by about a third every year. At current pace the world has three years of physical silver runway. The metal matters for solar photovoltaic production, for any space-based data center buildout, and for the traditional industrial and jewelry demand that has no substitute. The squeeze on silver is not a 2035 problem. It is a 2028 problem.
The electrification narrative also has a land constraint that rarely makes the slides. A one-gigawatt all-solar AI data center, factoring in the 20% capacity factor of solar, requires roughly five gigawatts of installed solar capacity, occupying about 35,000 acres — a footprint larger than San Francisco. Scaling solar is not a near-term solution to AI power demand, not because the sun does not shine, but because the permitting, interconnection, and land-assembly timelines push the build into the same decade as the data center it is supposed to feed. "We were swimming in natural gas in this country," Dreyfus said. "Nuclear — we can't really build it. We can't even build the containment vessels in this country. The Koreans can do it but we can't do that here." The energy bottleneck is not a resource problem. It is a construction problem.
The construction problem, in turn, is a labor problem. Dreyfus calls the craft labor shortage the single biggest bottleneck to executing the reshoring and electrification buildout. The blue-collar workers displaced by the 2000s offshoring wave did not sit around waiting to be called back. Many left the trades entirely. The cohort that stayed, or that has come through accelerated apprenticeship programs like Quanta's, is now commanding starting salaries of $150,000 out of high school — a wage inversion that the policy class has not yet metabolized. You cannot run a re-industrialization supercycle with a labor force that does not exist, and training the labor force takes the same decade that the mines and the gigawatts are supposed to come online.
Why hard assets, why now
Underneath the supply-demand arithmetic sits a macro argument that the policy response does not address. The US carries roughly $40 trillion in government debt, growing at about $2.5 trillion a year, on top of approximately $100 trillion in present-value unfunded social liabilities across Medicare, Medicaid, Social Security, and federal pensions — another $2.5 trillion a year. Total tax receipts run around $5.5 trillion annually. The arithmetic guarantees some combination of inflation, default, or financial repression over the coming cycle. The 1970s precedent — when the dollar lost 70% of its purchasing power and hard assets dramatically outperformed — is, in Dreyfus's framing, the most relevant template. "Commodity cycles typically last 15 years and have multiple hundreds of percent of upside," he said. "We're only a few years into this." His base case is that copper at least doubles from current levels. He has watched molybdenum go from a dollar a pound to $33, so a 2x in copper would be, in his telling, no big deal.
The policy response — equity, permits, offtakes with floor prices — is the most aggressive industrial policy the US has run since at least the 1980s. It signals that Washington has internalized the lesson of the 2010s, when it watched China consolidate the processing layer of nearly every critical mineral and concluded that the next conflict, the next pandemic, or the next trade war could close the choke points overnight. The three-pillar deal is the kind of instrument that only gets built when a Ford production line genuinely stops. Whether it scales fast enough to matter inside this decade is a different question. The permitting clock, the labor clock, and the geological clock are all running in years, not months.
Dreyfus ended his appearance with what amounted to a forward indicator. "If you want to look around the corner and see the next bottleneck coming, I strongly urge you to look at copper." It is the metal that electrification, defense, AI, and grid hardening all need simultaneously, and the metal whose supply curve is the hardest to bend. The world is heading into a decade in which the limiting factor on the energy transition, the AI buildout, and the reshoring of strategic industry will be the willingness of someone, somewhere, to dig a very large hole in the ground and wait the better part of a decade for it to start producing. The US has decided, in the past eighteen months, to be that someone. The clock is already running.
Wire provenance
This editorial synthesis draws on the following public wire/social posts:
- https://www.youtube.com/watch?v=xTO1aQ_m44I