History, it turns out, has a habit of rhyming in the most expensive ways possible.
"The shale boom was arguably the most notorious 'growth at all costs' capex cycle in the modern era, where energy industry-wide capex reached 110–120% of cash flow at its peak," Jeff Currie says. "So for technology spending to reach energy industry levels should raise a lot of questions."
The shale producers poured capital into the ground at a rate the economics never justified. A price crash followed, erasing $2.6 trillion in equity. The question now is whether the AI capex cycle risk is being written into the same script.
Deja Boom
Big Tech's current investment wave has a structure that Currie finds strikingly familiar.
Capital is flooding into chips and data centers. The confidence underpinning that spending, that AI energy demand and compute prices will stabilize at levels justifying the outlay, echoes the same conviction shale producers held about $100-per-barrel oil.
"Confidence in future AI computing prices stabilizing around the $1- to $2-per-hour range echoes the same confidence that the US shale producers had in $100/bbl oil that drove their spending far above cash flow," Currie noted.
The financing architecture maps across, too.
US oil producers once kept drilling debt on their balance sheets while offloading pipeline capex onto special-purpose vehicles. Today's hyperscalers are doing something structurally identical, using SPV arrangements to fund data center buildouts off-balance-sheet.
"Big Tech AI appears to be using the exact same playbook that the energy industry used as these arrangements clearly rhyme with today's AI datacenter SPV arrangements," Currie said.
The Pipe, Not the Pump
In every great infrastructure build-out, the companies doing the building rarely capture the full upside. The shale drillers who bet on $100 oil watched $2.6 trillion in equity evaporate across the energy sector. The picks-and-shovels suppliers often outlast the speculators.
For the current AI infrastructure commodities cycle, Currie's framework points in a clear direction, toward the physical world that the digital one depends on.
Natural gas AI data centers represent one of the most direct expressions of this thesis. The energy required to run large-scale compute comes from pipelines, power plants, and transmission infrastructure. Copper demand artificial intelligence tells a similar story. Every data center, every EV charging network, every grid upgrade requires copper at a scale that existing supply chains were never designed to meet.
The numbers are already biting.
The refined copper market is forecast to shift from surplus to a deficit of at least 150,000 tonnes in 2026, and with mine development timelines averaging 20 to 30 years in developed markets, new supply simply cannot respond quickly enough.
S&P Global projects the gap will widen substantially over the longer term, with global copper demand rising from 28 million metric tons in 2025 to 42 million metric tons by 2040, a 50% surge driven substantially by AI infrastructure, electrification, and data center expansion.
Oil supply chains have been substantially underinvested for years, with no major non-OPEC supply surge visible after 2026. The physical constraints are real, and they bite whether or not the AI companies themselves survive the cycle.
When Software Runs into a Wall Made of Metal
Energy's weight in the S&P 500 has fallen to roughly 3%, down from 25% in the 1970s.
At the same time, technology now accounts for approximately 53% of the index. That concentration, Currie argues, leaves most institutional portfolios acutely exposed if capital begins rotating toward real assets.
And the scale of the potential rotation matters: the market capitalisation sitting in the MAG7 alone dwarfs the entire global mining and energy sector. Even a modest reallocation would move commodity markets in ways the price signals haven't fully priced.
The concept Jeff Currie has developed to capture this is HALO: Heavy Asset Low Obsolescence. The framing is deliberate. These are companies and sectors built around physical infrastructure: energy, metals, mining, pipelines, railroads, utilities. They share one defining characteristic: software cannot disrupt them. A language model cannot replace a copper mine.
The HALO thesis requires only the recognition that every inflection point over the past fifty years has, eventually, triggered a rotation from asset-light to asset-heavy sectors.
The timing is never obvious. The direction, historically, has been.
The Old Economy Gets Its Revenge
The shale analogy was never meant to be a prediction of collapse. It was a warning about where the cycle ends up, and who is left holding value when it does.
Currie has framed this broader theme as the "revenge of the old economy," a phrase worth sitting with. The suggestion is that the assets dismissed as legacy infrastructure, as stranded, as irrelevant to a software-defined future, are quietly reasserting themselves. Geopolitical fragmentation is raising security premiums on commodities. The push toward electrification is increasing metals demand structurally. Nations are stockpiling. Supply chains built for efficiency are being rebuilt for resilience.
None of that requires AI to fail.
The shale producers didn't all fail either. Plenty of the underlying resource was real. The mistake was in the capital structure, in spending at 110% to 120% of cash flow in the belief that prices would justify it forever.
The companies that survived, and ultimately prospered, were often not the ones drilling the wells. They were the ones who owned the pipelines, the processing capacity, the hard infrastructure that the whole edifice required.
Jeff Currie's case, stripped back to its core, is that the same logic applies today. The AI boom is real. The physical world it runs on is scarcer than the market currently acknowledges, and the investors best positioned for what comes next may be the ones who stopped staring at the software and started looking at what sits underneath it.
Frequently Asked Questions
How will AI affect commodity markets?
AI is driving a structural increase in demand for physical commodities, particularly energy, copper, and metals, through the mass buildout of data centers, power infrastructure, and grid expansion. That demand is colliding with years of chronic underinvestment in supply, creating the conditions for sustained upward price pressure. The broader effect, as Currie frames it, is a capital rotation away from asset-light tech toward hard, physical assets that the digital economy depends on.
Is the AI investment boom creating a commodity supercycle?
Currie's case is that the conditions for a commodity supercycle are already in place: chronic underinvestment in traditional supply chains, rising geopolitical security premiums, structural demand shifts from electrification and AI, and the early stages of a capital rotation from financial assets into real ones. A full supercycle requires that rotation to accelerate. The AI buildout, by dramatically increasing physical commodity demand while supply remains constrained, is one of the more credible catalysts for that shift.
What commodities does the AI boom require?
The most direct requirements are energy, particularly natural gas to power data centers, and copper, which is embedded in every layer of data center infrastructure, from power distribution and cooling to networking and grid connection. Besides those two, the broader AI buildout increases demand for metals including silver, as well as the infrastructure assets that sit around the energy system: pipelines, utilities, and transmission capacity.
Is AI spending heading for a crash like shale?
Currie stops short of predicting a crash, but the structural warning is clear. The shale boom didn't fail because the oil was never there; it failed because capital was deployed at a rate the economics couldn't justify, underpinned by financing structures that offloaded risk rather than eliminated it. The same SPV arrangements, the same land-grab psychology, and the same confidence in stable future prices are all present in the current AI cycle. Whether that ends in a correction depends largely on whether AI revenues grow fast enough to justify spending that is, for several of the largest players, already running at 45 to 57% of revenue.
Why does AI increase demand for copper and natural gas?
Data centers are among the most energy-intensive facilities ever built, requiring continuous, large-scale power supply, the majority of which currently comes from natural gas. Copper is the connective material running through every part of that infrastructure: wiring, cooling systems, power distribution, transformers, and the grid upgrades required to deliver electricity to the sites in the first place. As data centers scale in both number and individual capacity, the copper and natural gas required to build and run them scales with them, and unlike software, neither commodity can be updated, compressed, or made more efficient by writing better code.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
