We Are Financing the AI Economy Before It Exists
The artificial-intelligence boom contains a curious inversion.
Usually, a technology spreads through the economy, proves its value, and then attracts the infrastructure, political support, and institutional change required to sustain it.
With AI, much of that sequence is running backward.
Governments are redesigning energy policy around future data centers. Companies are restructuring workforces around expected automation. Universities are rebuilding education around anticipated labor-market disruption. Investors are financing chips, electrical generation, transmission lines, and server campuses on the assumption that machine intelligence will become embedded in nearly every form of economic activity.
Yet new evidence from the United Kingdom suggests that most businesses remain much closer to experimentation than transformation.
The share of firms using some form of AI has risen substantially. But only a small minority describe that use as extensive. Most are using a narrow collection of tools, often for routine efficiency rather than new products, new markets, or a fundamental reorganization of work.
That does not mean AI is hollow. The systems are improving rapidly, and certain industries are already being reshaped. But it does mean the economic story is running several chapters ahead of the evidence.
This matters because the costs of anticipation are not imaginary.
Communities are being asked to accept enormous data centers, new transmission infrastructure, heavy water demand, tax incentives, and potential increases in utility rates. Florida politicians are now proposing that large AI facilities provide their own electricity and water rather than shifting those expenses onto the public.
That proposal is emerging from the political right, while environmental groups and progressive communities have raised similar concerns elsewhere. The coalition is unusual because the burden is unusually concrete. Ideology becomes less tidy when a household opens its power bill.
The same anticipatory logic is appearing in employment.
Companies may cut workers not only because AI has already automated their jobs, but because executives expect competitors to automate first. A recent economic model describes this as an AI layoff trap: each firm receives the full benefit of reducing labor costs, while the resulting decline in consumer demand is spread across the economy.
The company may be acting rationally.
The system may still be behaving irrationally.
This is the central contradiction of the present AI transition. We are told that the technology is too important to slow down, while being asked to accept that its broad economic benefits may take years to arrive. Infrastructure must be built now. Workers must adapt now. Communities must surrender resources now. Regulators must remain flexible now.
The benefits, meanwhile, remain written largely in the future tense.
Public investment in transformative technology is not inherently a mistake. Electrical grids, railways, highways, telecommunications, and the internet all required societies to build ahead of immediate demand. AI may eventually justify enormous investment and deliver extraordinary advances in science, medicine, education, accessibility, and productivity.
But public investment is different from public submission.
A serious AI policy would require developers to bear the costs they create, share measurable gains, disclose realistic demand projections, protect communities from utility-price increases, and provide evidence before layoffs are described as technological necessity. It would build public research capacity rather than leave safety testing dependent on unstable institutions and private laboratories.
It would also preserve the right to revise the plan.
The danger is not merely that the AI boom could fail. The more plausible danger is that the technology succeeds unevenly while the public absorbs the infrastructure costs, workers absorb the transition costs, and a narrow group captures most of the resulting value.
We may indeed be constructing the foundation of a new economy.
But before pouring another trillion dollars of concrete, copper, silicon, and debt, society should be permitted to inspect the blueprint and ask a rather ordinary question:
Who owns the building once everyone has paid for it?