Everyone Is Watching AI Stocks. Who’s Watching What AI Needs?

Everyone is watching AI stocks. Far fewer are watching what AI actually needs to run, which is a physical supply chain that begins with a chatbot and ends in a copper mine: AI leads to energy, energy leads to copper and uranium, and all of it leads to infrastructure. As of August 2026, six experts interviewed on Wealthion argue that the investable story may sit further down that chain than the crowded names at the top. Here is the chain, one link at a time.

What does AI actually need to run?

Not just chips. Brett Rentmeester of WindRock Wealth, in his July interview, traced the money from the top down: hyperscaler capital spending on data centers, the physical byproduct of the AI boom, was "about 156 billion in 2023" and has climbed steeply since. And what does that spending actually buy? In his words, the data centers "need chips. They need energy. they need real estate," plus water and land. The AI trade most investors own is the frontier labs and the chipmakers; the trade fewer are watching is everything those data centers must consume to exist. Rentmeester calls it "there's this whole interconnected web," and the web extends far past the ticker symbols on the front page. Follow the chain in one line and it reads: AI to energy, energy to copper and uranium, and all of it to the infrastructure that ties them together. Each arrow is a demand transfer from the crowded top of the market to the under-owned bottom.

Why is energy the first bottleneck?

Because computation is, physically, converted electricity. Michael Green of Simplify Asset Management put it memorably in his May interview: "energy commodities are very much machine food." As more of the economy's power flows to data centers, the machines compete with humans for the same grid, and Green expects the response to run through one channel in particular: "that creates demand for the energy infrastructure and in particular I'd highlight areas like nuclear." Steven Feldman, in his June interview, states the ranking flatly: "Energy infrastructure has got to be the single biggest theme around AI," against a backdrop of a trillion dollars being spent on the technology. The first thing AI needs, before any metal, is power it can actually plug into.

Why does AI drive demand for copper and uranium?

Because the energy has to be generated and delivered, and that is a metals problem. Uranium is the fuel for the nuclear capacity Green points to. Copper is the nervous system: every data center, transmission line, and generator is wired with it. The catch is supply, and Rick Rule, in his June interview, explained why new copper is so hard to bring online. Describing a rare high-grade discovery of "1.25 or 1.3% copper," he noted the best remaining ground sits in politically difficult regions that have been "underexplored," because the probability of "discovering billion ton" porphyry deposits is far higher there than in safer jurisdictions investors prefer. The metal AI needs exists; getting it out of the ground is a decade-long problem, which is exactly what turns demand into price. Uranium tells a similar story: years of underinvestment left few new mines, so a demand surge from nuclear-powered data centers meets a supply base that cannot scale on the same timeline.

Is this just the commodity supercycle in disguise?

Partly, and the experts are candid about the overlap. The difference this article draws out is the mechanism: AI is the specific demand shock that activates the broader scarcity story covered in our commodities analysis. At the 2026 Rick Rule Symposium, captured in this panel, Grant Williams offered the cleanest way for an ordinary investor to think about it: you can "hedge the fact that your electricity will be more expensive, that your automobile will be more expensive as a co as a consequence of the increased copper price" by owning the inputs, "by investing in the very things that you consume." In that framing, the AI supply chain is not a speculative bet on robots; it is a hedge on the rising cost of the physical economy AI is straining.

What could break the chain?

Two honest risks, both from within these interviews. The first is efficiency. Green, despite his energy-demand thesis, cautions that "there are extraordinary breakthroughs on efficiency within those data centers that mean many of the forecasts that people are making are overstated." If AI learns to do more with less power, the downstream demand for energy and metals softens. The second is monetization. Anthony Scaramucci of SkyBridge Capital, in his June interview, is bullish on the technology over a decade but names the open question directly: "there's a tremendous amount of capital expenditure that has to go into AI," and "there's a big question about can they monetize" it. If the AI business model disappoints, the capex that drives the whole physical chain slows with it. The supply-chain thesis depends on the buildout continuing.

How are these experts positioning?

Each playbook is the named expert's own. Rentmeester looks past the mega-cap names toward the physically enabled layer, including energy, industrial metals, and infrastructure plays in both public and private markets. Green favors energy infrastructure and nuclear as the durable demand sink. Feldman screens for scarcity and secure local supply, warning that the real constraint is often "scarcity is driving decisions" through the ability to secure supply, not just its global quantity, which is why "it could be a global glut and a local local scarcity" at once. Rule hunts the copper and uranium producers whose supply cannot be conjured quickly. And the symposium's consistent message is to own the inputs you consume. None of them is buying "AI" as a slogan; they are buying the physical things AI cannot run without. That is the whole inversion of this article: the crowd is positioned at the visible top of the chain, while the experts interviewed on Wealthion are positioned at the constrained bottom of it.

FAQ: What AI Needs in Brief

What does AI need to run besides chips? Enormous amounts of electricity, plus the copper to wire it, the uranium and natural gas to generate it, water for cooling, land, and the data-center infrastructure itself. Chips are one link in a long physical chain.

Why is AI driving up demand for energy? Data centers convert electricity into computation at scale. As Michael Green frames it, energy is machine food, and machines increasingly compete with households for the same power, pushing demand toward new generation, especially nuclear.

How does AI affect copper and uranium? Copper wires every data center, grid connection, and generator; uranium fuels the nuclear capacity being built to power them. Both face slow, expensive supply growth, which is what can turn AI demand into higher prices.

Is the AI supply chain the same as the commodity supercycle? They overlap. AI is a specific, powerful demand driver that activates the broader under-supplied commodity story, giving the supercycle thesis a concrete catalyst rather than a general one.

What could go wrong with the AI infrastructure trade? Two things named by these experts: efficiency breakthroughs that cut AI's power needs (Michael Green), and a failure of AI companies to monetize, which would slow the capital spending driving the whole chain (Anthony Scaramucci).

Which experts and interviews does this article reference? Six Wealthion interviews from May to July 2026: Brett Rentmeester on AI's $920 billion gamble, Michael Green on energy as machine food, Steven Feldman on the AI resource rush, Rick Rule on copper supply, the 2026 Rick Rule Symposium panel, and Anthony Scaramucci on AI and monetization.

Wealthion editorial content is for informational purposes only and is not investment advice. The views quoted belong to the named guests. If you want a professional read on how the AI infrastructure theme fits your own portfolio, you can request a free portfolio review at wealthion.com/free.