The AI Bubble Isn't in the Chips — It's in the Copper and the Loans

The AI Bubble Isn't in the Chips. It's in the Loans — and the Copper.

For three years, the AI trade has been an equity story. Nvidia. The hyperscalers. The Magnificent Seven. When investors worried, they worried about the valuation of a handful of chip stocks.

This month, the story changed.

Bain & Company surveyed 951 large companies and reached a conclusion that should stop every allocator cold: "The technology worked. The value didn't arrive." A National Bureau of Economic Research study found roughly 90% of firms reporting no measurable productivity gain from AI — even as most plan to spend more. Fortune is running "1999 again." Bloomberg is asking what happens if the bubble bursts.

The consensus is finally pricing risk. But it is still watching the wrong layer.

The danger was never really in the chip stocks everyone follows. It sits in two layers almost nobody is pricing.

Layer one: the physical wall

A single one-gigawatt AI data center can require up to 50,000 tons of copper — for power, grounding, cooling. Against that demand sits a copper market already in a structural deficit, with transformer lead times stretching from two years to four and power-interconnection queues that do not clear on an earnings-call timeline.

We flagged this in May, while the consensus was still euphoric: $725 billion committed to AI, and the copper that buildout depends on hasn't been mined. In June, the analysts confirmed the deficit. The buildout cannot be poured fast enough — and that is not a sentiment problem. It is a physics problem.

Layer two: the credit

Here is what turns a tech-sector story into a systemic one. AI capex is increasingly debt-financed — through private credit, a market now facing what allocators openly call its "first big test since 2008." The European Central Bank's latest Financial Stability Review flagged the sharpest tightening of corporate credit standards since 2023.

Follow the chain: record spend → returns that haven't arrived → the debt that financed the spend gets exposed → private-credit stress → and stress in private credit does not stay in technology. It travels through the institutions that hold it.

Why the gap matters

None of this means the technology is fake. In 1999, the internet was not fake either — it went on to remake the global economy. But the gap between what was spent and what came back was wide enough that most of the investors who funded the build did not survive to see the payoff.

That is the shape of the risk now. The consensus is priced for a clean story: AI works, therefore the spend is justified, therefore the debt is fine. The physical layer and the credit layer say the path from here is not clean.

By the time "AI bubble" is the headline — and this month, it became the headline — the repricing is already underway. Capital that is certain moves first; prices follow.

The question for portfolios

For anyone exposed to AI, technology credit, or the broader risk complex, the question is no longer "is AI real?" It is: are you priced for the demand — or for the copper and the loans that demand actually runs on?

At VestAI we map that gap — the physical and financial layers underneath the narrative, visible while they are still forming rather than after the projects slip.

What's your read — is this 1999, or a payoff that's simply late?

#AI #privatecredit #commodities #riskmanagement #markets