A sophisticated take on the AI bubble argument

GeometricInvestor has a longish analysis of money flow and investment in the AI business, The AI Capex Ledger: Who Pays, Who Earns, and What the Bond Market Is Missing, June 12,2026.

The opening paragraphs:

The AI debate is stuck on the wrong question.

The question is not whether AI is a bubble. Nor is it whether NVIDIA is expensive, whether Michael Burry is early again, or whether productivity gains will eventually lower inflation. The better question is an accounting one: who needs to earn what return for the AI capex cycle to make sense?

At the bottom of the stack, GPU, HBM, networking, power, and cooling suppliers are already earning. The capex is real. The checks have cleared. That was the first phase of the trade.

The harder question sits one layer above. Hyperscalers and neoclouds are converting capital into compute. Compute becomes tokens. Tokens must become revenue. Revenue must become gross profit after depreciation, power, financing, and model costs. And finally, the buyers of those tokens must earn a return high enough to keep spending.

Only then does AI become a true macro productivity shock rather than a capital-spending boom with better branding.

Then comes the long analysis, followed by these concluding remarks:

The AI cycle will not be resolved by asking whether GPUs are expensive or whether chatbots are useful. It will be resolved by a chain of returns.

Can infrastructure suppliers earn margins without oversupply?

Can hyperscalers and neoclouds sell enough tokens to cover depreciation, power, financing, and obsolescence?

Can enterprises earn more from those tokens than they spend?

Can the economy convert those firm-level returns into productivity growth?

And if it can — does the bond market understand that higher productivity may mean a higher neutral rate, not just lower inflation?

That is the real AI macro debate. Each layer has a hurdle, and each hurdle has a date with evidence. The next time an AI headline crosses your screen, skip the bubble question and ask the ledger question: which layer is this, and what return does it need?

The framework carries its own falsifier, as it should: monetizable AI revenue clearing the hurdle band for years while aggregate productivity and corporate operating leverage outside the supplier base stay flat. That would mean buyers funding sellers indefinitely without a return — the one thing this piece says cannot last.

The first phase was about capex. The next phase is about returns.

H/t Tyler Cowen.