The $166bn Tariff Rebate

For roughly a year, the largest single change in the effective tax burden on American business was not written by the tax-writing committees, scored by the Joint Committee on Taxation, or debated on the floor of either chamber. It arrived by executive order, was collected at the border, and was booked by companies as cost of goods sold.

On 20 February 2026, the Supreme Court held that it should never have been collected at all. In Learning Resources, Inc. v. Trump, decided 6–3 with Chief Justice John Roberts writing for a majority that included Justices Sotomayor, Kagan, Gorsuch, Barrett and Jackson, the Court concluded that the International Emergency Economic Powers Act does not confer on the President the authority to impose tariffs. The so-called "reciprocal" tariffs first announced in April 2025, and the separate trafficking-related duties on Canada, Mexico and China, fell together.

What followed has been one of the largest and strangest fiscal events in modern American administration: a reverse tax collection running to well over a hundred billion dollars, processed through a system that did not exist six months ago, distributed to a population of claimants that no one had ever needed to enumerate, and taxable — at least in part — on receipt.

The refund is not a footnote to the tariff story. It is where the tariff story becomes a tax story, and where the questions that tariff policy has largely been allowed to avoid — who actually bears the cost, who is entitled to the benefit, what the fiscal accounting really looks like — become unavoidable, because someone now has to cut the cheques.

The size of the hole

The numbers are unusually well documented, because the Court of International Trade has required the government to account for them publicly on a rolling basis.

U.S. Customs and Border Protection has told the court that approximately $166 billion in duties was collected under the IEEPA authority. The Penn Wharton Budget Model, working from International Trade Commission and Treasury data, projected refunds of up to $175 billion and estimated that at the time of the ruling IEEPA duties were running at roughly $500 million per day and accounted for slightly more than half of all customs duties collected by the United States. Its blunt summary of the fiscal consequence was that, absent a replacement authority, future tariff revenue would fall by about half.

The claimant population is similarly large. More than 330,000 importers of record paid IEEPA duties across more than 53 million individual entries. For scale: in the whole of the preceding fiscal year, CBP processed roughly 338,000 entries involving tariff refunds of any kind. The agency has now processed IEEPA refunds on millions of entries in a matter of weeks.

Two decisions of the Court of International Trade — Atmus Filtration, Inc. v. United States in early March and Euro-Notions Florida, Inc. v. U.S. Customs and Border Protection in April — ordered CBP to refund IEEPA duties to all importers of record, not merely to the plaintiffs before the court. The CIT then partially stayed its own orders while CBP built the machinery to comply, requiring weekly status reports in the interim.

That machinery is the Consolidated Administration and Processing of Entries system, or CAPE, a new module inside CBP's existing Automated Commercial Environment. It went live on 20 April 2026 and is the exclusive electronic route for IEEPA refund claims. Refunds are paid with statutory interest, typically within 60 to 90 days of a claim being accepted, subject to compliance review.

By late June, CBP had authorised roughly $104 billion in refunds and disbursed about $71 billion. Trade-data trackers put the figure repaid by 10 July at approximately $86 billion, with substantially more accepted into the processing queue. Whatever the precise running total on any given day, the order of magnitude is settled: this is a transfer from the federal government to the private sector on the scale of a mid-sized stimulus programme, executed without a single vote in Congress.

A refund system with a door in it

CAPE was designed in phases, and the phases matter more than they sound, because they determine who gets paid and who litigates.

Phase 1, live since April, covers entries that had not been finally liquidated when the refund clock started, plus entries liquidated within the preceding 80 days. CBP's Executive Assistant Commissioner for Trade, Susan Thomas, testified in June that Phase 1 had accepted claims covering roughly $90 billion of the $166 billion total, with about $23 billion approved and transmitted to Treasury at that point. Phase 1 captures something in the region of 63 per cent of affected entries.

Phase 2, launched at the end of June, extended coverage to reconciliation entries and certain entries subject to antidumping and countervailing duty orders, subject to conditions. CBP estimated it would reach a further 2.8 million entries and roughly $28.7 billion, bringing combined coverage to about $130 billion of the $166 billion.

That leaves a residue of roughly $30 billion or more in "finally liquidated" entries — those liquidated more than 80 days before the refund period began. Under customs law, liquidation is the point at which the duty calculation on an entry becomes final and, ordinarily, unchallengeable. Phase 3, targeted for late July, was designed to address these entries. But the government's stated position is that Phase 3 refunds will be processed only for importers who have filed suit at the Court of International Trade — a group of roughly 4,000 plaintiffs.

On 3 June, the Department of Justice filed notices of appeal at the Federal Circuit challenging the CIT's universal refund orders. The government's expected argument is that those orders function as impermissible universal injunctions under the Supreme Court's 2025 decision in Trump v. CASA, and that importers who did not bring their own actions are not entitled to the benefit of relief obtained by others.

The plaintiffs' anticipated answer is that CASA does not govern here, for two reasons. First, the Court of International Trade is not an ordinary district court; its jurisdiction is statutorily tied to the uniform administration of customs duties, so an order requiring CBP to treat identically situated importers identically is arguably a core function rather than an overreach. Second, the usual objection to nationwide relief — that it settles national policy before the merits are resolved — has no purchase when the merits have already been decided by the Supreme Court. The question is not whether the tariffs were unlawful, but whether the trial court may require uniform implementation of a holding that they were.

There is a further, more technical argument: that the entire architecture of final liquidation presupposes duties lawfully imposed, and that where the imposing authority never existed, finality rules addressed to computational disputes should not bar recovery.

Both positions are serious, and the outcome is genuinely uncertain. What is not uncertain is the practical effect of the litigation posture in the meantime. The government has drawn a line between importers who sued and importers who did not. Counsel across the trade bar have converged on the same advice — file a protective action at the CIT while continuing to claim through CAPE — which means the refund process is generating its own secondary wave of litigation, and that the firms best able to absorb the cost of a precautionary lawsuit are the firms most likely to be made whole.

The refund is not free money

Here the trade story hands off to the tax code, and the handoff is not gentle.

Most importers treated IEEPA duties the way duties are ordinarily treated: as a cost of acquiring inventory, folded into cost of goods sold and deducted as that inventory was sold. Some capitalised duties paid on capital equipment into the basis of the asset and began depreciating them. A great many paid duties on goods bought from foreign affiliates, where the duty interacted with an intercompany transfer price.

Each of those treatments produces a different answer on the way back.

The governing principle for the most common case is the tax benefit rule: a recovery of an amount deducted in a prior year is generally income in the year of recovery, to the extent the earlier deduction produced a tax benefit. An importer that paid and deducted IEEPA duties on goods sold during 2025, reducing 2025 taxable income, will generally include the refund in 2026 gross income. An importer that was in a loss position, or otherwise did not obtain the full benefit of the deduction, may exclude some or all of the recovery.

Where the tariffed goods were still sitting in inventory at the end of 2025, the analysis differs: no deduction was taken, so there is no recovery to include. The refund instead reduces the carrying cost of that inventory — practically, a reduction to 2026 opening inventory — and the benefit emerges later, as lower cost of goods sold when the goods are eventually sold.

Where duties were capitalised into equipment basis, the refund requires a basis adjustment and, depending on the depreciation already claimed, may trigger recapture.

The transfer pricing dimension is the least settled. Where an importer paid duty on goods acquired from a related foreign supplier, the question of which entity in the group is economically entitled to the refund is not answered by the customs rules, which look only to the importer of record. Groups that treated the duty as a cost borne by the U.S. distributor in setting 2025 intercompany prices may find their 2026 transfer pricing affected, and the refund may cascade into entered value and other duties in ways that are still being worked through.

The aggregate effect is that a meaningful fraction of the refund will come back to Treasury as corporate income tax, on a lag, at whatever the taxpayer's effective rate turns out to be. That is a genuine offset to the fiscal cost — but it is an uneven one. It is largest for profitable importers that fully deducted in 2025, smallest for loss-makers, and deferred for those holding tariffed inventory. The refund is therefore not simply an unwinding of the tariff. It is a second redistribution, in a different direction, on a different timetable, with different winners.

Who actually paid the tariff?

This is the question the refund forces into the open, and it is the one on which reasonable people most sharply disagree.

The legal answer is unambiguous: the importer of record paid the duty, the government collected it unlawfully, and the importer is the party entitled to restitution. Refunding money to the person from whom it was taken is not a policy choice; it is the ordinary remedy.

The economic answer is less tidy. Empirical work on the 2018–2025 tariff episodes has generally found high pass-through of tariff costs into domestic prices, with published estimates commonly falling in a range of roughly 80 to 95 per cent. If that is broadly right, the great majority of the economic burden of the IEEPA tariffs was borne not by the importers now receiving refunds but by the businesses and households further down the chain who paid higher prices.

Those purchasers have no claim. There is no mechanism by which a consumer who paid more for a washing machine in 2025 can recover anything, and no serious proposal to create one. The result is a mismatch between economic incidence and legal entitlement that is unusual only in its scale — the same mismatch arises whenever an indirect tax is refunded — but which at $166 billion is large enough to be a distributional event in its own right.

Whether that mismatch constitutes a windfall depends on what one thinks firms do with the money. Some of it will be competed away: in markets with thin margins and mobile customers, a restored cost base tends to show up in prices, though usually more slowly on the way down than it did on the way up. Some will be retained. Apple disclosed a $2.2 billion tariff benefit in a single quarter and indicated it would direct the money towards U.S. manufacturing investment rather than price reductions — a defensible corporate decision, and also a clean illustration of why "the consumer paid, so the consumer should be repaid" is easier to assert than to operationalise.

There is a further wrinkle that is easy to miss. Firms that passed the tariff through in full booked the higher revenue and paid tax on the resulting margin. Firms that absorbed it took the hit to income. The tax system has already, imperfectly, sorted some of this. Layering a taxable refund on top produces outcomes that vary enormously by firm circumstance and are difficult to characterise as systematically fair or unfair.

The concentration problem

Even setting aside the consumer question, the refund is not landing evenly across the business population.

Analysis by the Cato Institute found that roughly 60 per cent of refund dollars traced to about 30 per cent of import entries — a concentration that follows mechanically from the fact that a small number of large importers account for a disproportionate share of high-value entries. Large firms also have the customs compliance infrastructure to file millions of CAPE declarations accurately, the working capital to wait out a 60-to-90-day processing cycle, and the legal budget to file a protective CIT action against the possibility that the Federal Circuit narrows the universal refund orders.

Smaller importers face the mirror image on each count. Many used brokers rather than in-house customs teams, hold entries in the categories that fell outside Phase 1, and face a real cost-benefit question about whether a protective filing is worth it for a five- or six-figure claim. Some have already been reported as pursuing claims for duties they did not themselves remit, which will generate its own compliance and enforcement problem for CBP.

None of this is anyone's design. It is what happens when a $166 billion remediation is administered through a customs infrastructure built for a fraction of the volume, on a compressed timetable, under judicial supervision. But the predictable result is that the refund tilts, at the margin, towards scale.

The fiscal picture

The refunds have been large enough to be visible in the headline federal numbers. The nine-month FY2026 deficit came in around $1.37 trillion, roughly 2 per cent wider than at the same point the prior year, with the refund wave a principal driver of the widening.

That figure understates the underlying change in two directions at once. It overstates the permanent cost, because a portion of the refunds will return as corporate income tax under the rules described above, and because interest paid on refunds is itself deductible-and-taxable in the hands of recipients. It understates the ongoing revenue effect, because the refunds are a one-off return of past collections whereas the loss of the IEEPA authority removed roughly half of the prospective customs duty base.

The administration moved to close that prospective gap within hours of the ruling, reimposing a flat 10 per cent tariff under Section 122 of the Trade Act of 1974. Section 122 is a balance-of-payments authority with a hard 150-day statutory life, and it did not survive its term intact: the Court of International Trade held the Section 122 tariffs invalid in May, and the authority lapsed by its own terms on 24 July 2026.

The replacement has been assembled from the two authorities that have proved durable. The U.S. Trade Representative concluded a set of Section 301 investigations and imposed new tariffs of 10 or 12.5 per cent on imports from some 60 trading partners, with product-specific exemptions, on grounds relating to the failure to prohibit and enforce against goods produced with forced labour. Section 232 duties — 50 per cent on steel and aluminium for most origins, 25 per cent for the United Kingdom, 25 per cent on copper — remain, with further sectoral action on pharmaceuticals and medical devices phasing in.

The trade-policy substance is therefore substantially preserved. The legal architecture is entirely different: where IEEPA offered a single instrument applied broadly and quickly, Sections 232 and 301 require investigations, findings, product-level determinations and notice-and-comment procedure. For businesses, that means the effective rate is now a function of origin, HTS code and exemption status rather than a headline percentage — more durable, considerably harder to plan around, and far more expensive to comply with.

What the episode actually decided

It is tempting to read Learning Resources as a case about tariffs. It is more usefully read as a case about which branch levies taxes.

Tariffs are taxes. They are collected from domestic parties, they are borne substantially by domestic purchasers, and they raise revenue at a scale comparable to significant provisions of the income tax code. For most of the past decade they have nonetheless been treated, in practice and in political discourse, as an instrument of foreign policy rather than fiscal policy — set by executive action, adjusted without scoring, and largely insulated from the procedural apparatus that surrounds every other federal revenue measure.

The Court's holding restores the formal position that the power to impose tariff measures rests with Congress. What it has not resolved is the practical one. Congress has delegated tariff authority extensively and repeatedly, and the post-IEEPA regime demonstrates that the remaining delegations are ample enough to reconstruct most of the policy, if more slowly and with more paperwork. The constraint that bound in February was the absence of a statutory basis, not the absence of executive will.

There are two readings of what follows, and both have serious adherents.

On one, this is the separation of powers working as designed, slowly. An unlawful exercise of power was struck down, the money is being returned with interest, and the executive has been pushed back into authorities that carry procedural discipline — investigations, findings, comment periods — which will make future tariff policy more deliberate and more contestable.

On the other, the episode demonstrates how little the formal allocation constrains. The tariffs ran for a full year, reshaped supply chains, and were substantially replaced within five months of being invalidated. The remedy arrived long after the economic effects had been absorbed, and it reaches only the party that remitted the tax, not the parties that bore it. On this reading the case is a lesson in the limits of after-the-fact judicial correction of fiscal measures — the money can be given back, but the year cannot.

What to watch

Three things will determine how this settles.

The first is the Federal Circuit. If the government prevails on the CASA argument, roughly $30 billion of finally liquidated entries becomes available only to the 4,000 or so importers who sued, and the refund becomes explicitly two-tier. If the plaintiffs prevail, the universal orders stand and the remaining phases proceed. A further petition to the Supreme Court cannot be excluded, and the pool of finally liquidated entries grows with every passing month of appellate briefing.

The second is the 2026 filing season. This is when the tax treatment stops being an advisory question and starts producing return positions at scale, on facts — inventory turns, loss years, intercompany pricing, basis in capitalised duties — that vary company by company. Expect controversy, and expect pressure on Treasury and the IRS for guidance in an environment where enforcement capacity has been materially reduced.

The third is Congress. The Court has handed the tariff power back. Whether the legislature does anything with it — by narrowing existing delegations, by imposing scoring or sunset requirements on executive tariff action, or by doing nothing at all and allowing the Section 232 and 301 architecture to become the permanent settlement — is the question that determines whether February 2026 was a turning point or an interruption.

For now, the most concrete legacy of the largest unlegislated tax increase in recent American history is a queue: 330,000 importers, 53 million entries, and a refund system built in ninety days, working through a hundred and sixty-six billion dollars that should never have been collected — and a tax bill, arriving next spring, on the money coming back.

This article is analysis and commentary, not tax, legal or investment advice.

The Hundred-Billion-Dollar Rebate: What happens when the government has to give back a tax it never had the power to collect