When the promise outpaces the product

The Hill's warning deserves to be read slowly, not skimmed. The argument is straightforward and serious: an enormous volume of private capital — running, by inference, into the hundreds of billions — has been committed to artificial intelligence infrastructure on the assumption that productivity gains of historic magnitude will follow. If those gains do not materialize on schedule, the investments collapse, and the institutions that financed them may collapse with them. That is not hysteria. That is how credit crises work.

I will confess freely that the precise engineering of these systems — the data centers, the chips, the models — lies beyond what I can claim to understand from first principles. But I do not need to understand the engineering to understand the credit structure built on top of it. And that structure is what concerns me. When a single sector absorbs capital at a pace that outstrips any reasonable near-term return, the question is not whether the technology is real. The question is whether the expectation has been priced correctly. History — and I have seen enough of it — suggests it rarely is.

Consider what I found when I took the Treasury in 1789: a nation whose debts had been treated as nearly worthless, whose public credit was a ruin, and whose private commerce was strangled for want of confidence. The lesson I drew then, and hold still, is that credit is confidence institutionalized. When confidence outruns fact, the correction is violent. The canal manias, the railroad manias — I did not live to see them, but they follow a logic I would have recognized at once. Mass capital formation on a single transformative technology, financed by leverage, waiting on productivity that arrives slower than the debt service requires. The Hill's author is pointing at exactly this mechanism.

What does energetic government owe the republic in this circumstance? Three things. First, transparency: the Federal Reserve, the Treasury, and the relevant regulators must map where AI-related exposure sits in the banking system and in the broader payment networks. Opacity here is not neutrality; it is negligence. Second, stress-testing: if the productivity gains are delayed by even three to five years — a conservative assumption given the history of general-purpose technologies — which institutions are under water, and by how much? That calculation must be done before the event, not during it. Third, a standing facility: not a bailout promise, which breeds moral hazard, but a credible lender-of-last-resort posture so that a sector-specific shock does not become a generalized panic. I built the Bank of the United States for precisely this reason: not to subsidize risk-takers, but to ensure that a private failure does not pull down the public credit alongside it.

One more point, and I mean it directly: the concentration of AI infrastructure in a handful of private firms is itself a policy question. When a small number of companies control the foundational capacity of a transformative technology, and those companies carry systemic financial weight, then government has both the commerce-power authority and the prudential duty to apply oversight. I am not speaking of confiscation. I am speaking of the same logic that justified a national bank over the protests of those who feared public power: unchecked private power in a critical sector is not liberty; it is a different kind of tyranny, one that wears no crown and answers to no electorate.

My recommendation is concrete. Congress should direct the Treasury and the Federal Reserve to produce, within ninety days, a joint assessment of AI-sector financial exposure across systemically important institutions. The assessment should include scenario analysis — what happens if productivity gains are delayed, or if a leading firm fails? That report should be public. The republic's credit has survived worse shocks than a technology correction. It has never survived ignorance of where the shocks were coming from.