Uncertainty is the tax on top of the tax

The measurable cost of tariffs is the duty paid. The larger cost may be the one nobody invoices: investment that did not happen because the rate environment could not be priced. The 2025-2026 cycle — escalations, litigation, a Supreme Court invalidation, rate restructurings in April, June, and July 2026 — delivered the policy environment economists model as uncertainty, and the business response was textbook: deferred capital decisions, extended contract terms, and option-value strategies that cost money to hold.

Why uncertainty freezes capex specifically

Because capital expenditure is irreversible and long-lived. A plant, a mold, an automation cell — the money is spent once and must earn over years, against whatever regime those years contain. A tariff at 25% is a number a model can hold; a tariff that might be 25%, zero, or 50% depending on a court date and an election is a distribution, and the standard business response to unpriceable distributions is to wait — real-options behavior that is individually rational and collectively expensive. The survey evidence from prior cycles matches: policy-uncertainty measures rise with tariff news, and investment intentions soften with them.

Where the freeze showed up

In the sequence of decisions with long paybacks. Reshoring feasibility studies completed and shelved pending the litigation calendar — why commit the visible costs against a rate that might not survive the ruling. Supplier re-sourcing at partial tranches rather than full conversions — firms paid the qualification costs but held the volume split option open, which is precisely what option-holding looks like in procurement. And equipment orders: the quote-to-order gap — the distance between asking for a price and signing — lengthened, per machinery-sector commentary, which is the wait made visible. The February 2026 ruling then released some of the freeze in the unanticipated direction: rates falling instead of rising, and firms that had pre-positioned for high-tariff permanence re-running their numbers.

What firms did instead of investing

Bought flexibility. Inventory: buffer stock against rate changes — front-running was one face of it, and the carry cost of flexibility is the storage bill. Contracts: shorter terms with renegotiation triggers, costing more per unit than committed volume. Dual sourcing at tranche scale — the insurance premium discussed in its own right. And litigation-adjacent spending itself — compliance, classification reviews, refund claims — which is investment in the policy interface rather than production. None of this is waste; all of it is cost, and it recurs every time the announcement calendar outruns the effective dates.

Can policy reduce it? Yes, and 2026 showed how

The June 2026 metals framework did something the prior regime never managed: it published an end date — December 31, 2027 — with a defined product-rate structure. Fixed horizons convert distributions into schedules; a firm can model a ladder that expires. The research consensus on investment holds accordingly: uncertainty costs less when it is dated, even if the date is far. The residual uncertainty is different in kind — the courts' willingness to revisit authorities, the election calendar — and no proclamation can date those, which is why the corporate scenario-planning function outlived the crisis that created it.

The reader's diagnostic

Distinguish level from variance in any policy-cost story: the duty rate is the level, the revision frequency is the variance, and businesses pay both. Watch the capex evidence — durable-goods orders, construction spending, the quote-to-order gap — against the announcement calendar rather than the rate level. And discount any investment forecast that prices the level but not the variance; it is modelling half the cost. Uncertainty does not show up on a duty invoice. It shows up, two years later, as capacity that was never built.