Tariff Refunds, the $40 Trillion Mark, and the Deadline Most Owners Haven't Calendared

If your business imported anything between February 2025 and February 2026, money is owed to you. Roughly $166 billion was collected under tariffs the Supreme Court held unlawful in February,¹² and about $100 billion of it has been certified for payment so far.⁴

The part that surprises people: nobody is going to send it to you.

Refunds are not automatic. They require an affirmative filing, in a system Customs built for this purpose, by the importer of record — who may not be you. $1.6 billion has been withheld for nothing more exotic than an out-of-date bank account.³ And for the oldest entries, the window to go to court closes in February 2027.⁵

Meanwhile the same refund wave helped push the national debt past $40 trillion for the first time, on August 18.⁶ The two facts are related, and both get written about badly — the first as a windfall, the second as a catastrophe. Neither framing survives contact with the numbers.

What Actually Happened, and What Replaced It

On February 20, 2026, in Learning Resources, Inc. v. Trump, the Supreme Court held 6–3 that the International Emergency Economic Powers Act does not authorize the President to impose tariffs.¹ Chief Justice Roberts wrote that IEEPA's list of powers — investigate, block, regulate, direct, nullify, void, prevent, prohibit — contains no mention of tariffs or duties, and that the government “cannot identify any statute in which the power to regulate includes the power to tax.”

That struck down both the trafficking tariffs and the reciprocal “Liberation Day” tariffs. It left every other tariff authority untouched.

The remedial questions went to the Court of International Trade, which on March 4 ordered Customs to reliquidate unliquidated entries without the IEEPA duties — and the government has since appealed the portion of that order covering importers who didn't sue.⁷

And tariffs themselves did not go away. They moved statutes.

• Section 232 metals tariffs expanded in April: 50% on primary steel, aluminum, and copper, 25% on many derivatives, now applied to the entire customs value of a derivative product rather than just its metal content.⁸
• Section 301 took effect July 24 at 10% to 12.5% across roughly 60 economies.⁹
• Section 338 of the Tariff Act of 1930 — never used in its 96-year history — was invoked against Canada at 50% on roughly $20 billion of goods, including motor vehicles, dairy, and alcohol. USMCA-originating goods are not exempt. Canada has announced dollar-for-dollar retaliation effective September 8, targeting steel, dairy, appliances, agricultural equipment, pulp and paper, and electronics; the total had not been published as of late August.¹⁰

The average effective tariff rate was roughly 16% the day before the ruling, 9.1% the day after, and is about 11% today.¹¹ Section 301 and 338 are procedurally grounded in agency investigations and statutory findings — considerably more durable than the emergency authority that just failed. Plan on tariffs as infrastructure, not as a shock to wait out.

The Refund Is Not Automatic, and There Is a Clock

Customs built a system called CAPE — Consolidated Administration and Processing of Entries — inside its existing ACE portal.¹² The importer of record, or the licensed broker who filed the entries, submits a declaration listing entry numbers. Customs recalculates, reliquidates, and pays by ACH with statutory interest.

Four things about it are worth knowing before your controller tells you it's handled.

It requires you to come forward. Customs is not scanning its own records and cutting checks. As one law firm put it, the design “relies on importers to come forward and submit evidence concerning entries.” Sophisticated importers with customs counsel filed in April. Others haven't filed at all.

Post Summary Corrections are the wrong tool. Customs said explicitly that PSCs should not be filed for unliquidated IEEPA entries. CAPE is the exclusive administrative route, and filing the wrong thing can slow you down rather than speed you up.

Your bank details have to be current. Customs went electronic-only for refunds effective February 6, 2026.¹³ An importer not enrolled in ACH refund through the ACE portal has its refund rejected. That is the entire explanation for the $1.6 billion sitting undelivered as of July 31 — not litigation, not backlog, bad account information.

And the entries that are already finally liquidated are the hard cases. Customs' authority to voluntarily reliquidate runs only 90 days past liquidation, and the window to protest a liquidation is 180 days.¹⁴ Past that, an importer may need a court order. The government has appealed the ruling that would give non-litigants relief, arguing it amounts to an impermissible universal injunction — so whether you have to sue is genuinely unresolved.

Timing on everything else is manageable: acceptance runs about ten days from submission, and refunds land roughly 60 to 90 days after that — call it 70 to 100 days end to end. Interest accrues from the original deposit date through reliquidation, at the quarterly rate Customs publishes. In the third quarter of 2026 that was 7% for non-corporate filers and 6% for corporations — a full point less for a C corporation on the same dollars.¹⁵

The Refund Goes to the Importer of Record — Which May Not Be Whoever Paid

Customs pays the importer of record. Full stop. If your goods came in under a freight forwarder's or a supplier's or an express carrier's IOR number, the refund lands in their account, not yours.

There is exactly one administrative lever, and it is narrower than it sounds: CBP Form 4811 lets the importer of record direct where a refund is sent. It has to be in place before the CAPE declaration is filed — afterward is too late — and it changes the delivery address, not who owns the money.¹⁶ Ownership is still a contract question.

Everything else is contract law and state law, and Customs has issued no guidance on pass-through obligations at all.

The express carriers are remitting, because they itemized duties as separate line items on invoices and have no argument that the charge was theirs. UPS has reclaimed roughly $5 billion and its CEO's public position is that the company is “just a pass-through.” FedEx has about $800 million and built a customer lookup tool. DHL filed on nearly all shipments where it was the importer.¹⁷

Most retailers are not, because they folded tariff costs into general pricing rather than itemizing them. In a single week in August, major retailers disclosed roughly $5.3 billion of refunds: Walmart $2.9 billion, Target $994 million, Home Depot $730 million — of which $685 million went straight into reducing cost of goods sold — TJX $331 million, Ross $253 million, Lowe's $80 million.¹⁸ Walmart's CEO said the intent was to “deploy much of that back into price.” Others applied it to margin. Amazon is refunding customers only in the “limited set of circumstances” where the charge is directly traceable, because the company or its sellers absorbed most of the tariff rather than passing it through.

That last point is the one to internalize, and it cuts both ways. Where the tariff was absorbed into price, the refund is properly the importer's margin recovery and there is no downstream claim. Where it appeared as a separately itemized surcharge, a downstream buyer has a real argument — breach of contract where the agreement obligates reimbursement of duties actually incurred, breach of the implied covenant, or unjust enrichment on the theory that the importer recovered twice.¹⁹

Consumer class actions are already filed against several large companies, and no court has ruled on the merits yet. Business-to-business claims — where the contracts are cleaner and the invoices itemized — are the more viable path and had not materially emerged as of August.

So there are two questions to run, not one. Did you pay a tariff surcharge as a line item to anyone? If yes, you may have a claim, and the records that prove it sit with your supplier. Did you charge one to anyone? If yes, you may have a contingent liability sitting against a refund your accountant has already booked as income.

The Tax Bill Nobody Modeled

This is where owners get hurt, and it is worth being precise, because there is no IRS guidance on tariff refunds. None.²⁰ Every position below rests on general authority applied by analogy, and each is cited.

The governing principle is the tax benefit rule under IRC §111, as construed in Hillsboro National Bank v. Commissioner: a recovery of a previously deducted amount is income only to the extent the original deduction produced a tax benefit.²¹ What that means in practice depends entirely on what you did with the duty when you paid it.

For most importers the refund is not “other income” — it is a reduction of cost of goods sold, which is precisely why it runs straight to gross margin and EBITDA. Hold that thought for the next section.

Three traps worth naming.

Timing can separate the income from the cash. For accrual-method taxpayers, income arrives when the all-events test fixes the right to recover. The closest authority is Rev. Rul. 2003-3, on state and local tax refunds, which puts inclusion at receipt or notice that the claim has been approved, whichever comes first — reasoning that turns on the reviewing authority's process being substantive rather than ministerial.²⁴ Whether Customs' acceptance of a CAPE declaration is that “approval” is unresolved. If it is, a claim accepted in late 2026 and paid in 2027 produces 2026 income with no 2026 cash.

The estimated tax safe harbor may not be there. A corporation with taxable income of $1 million or more in any of the three preceding years is a “large corporation” under §6655(g)(2) and loses the prior-year safe harbor except for its first installment, which is recaptured in the next one. It has to fund estimates on current-year tax or annualize.²⁵ Individual owners taking this through a pass-through need 110% of prior-year tax where prior-year AGI exceeded $150,000, under §6654(d)(1)(C).²⁶ A seven-figure refund can turn a routine estimated payment into an underpayment penalty.

Amending 2025 is not the mechanism. The 2025 deduction was correct when taken. This is a current-year inclusion regime, not a retroactive one.

And one genuine planning point: where customers hold enforceable pass-through rights, there is a serious argument under Commissioner v. Indianapolis Power & Light Co. that the importer never had complete dominion over the money and shouldn't include it at all — with the corresponding deduction for amounts owed back to customers governed by the economic performance rules of §461(h).²⁷ That is a conversation for your tax counsel, and it needs to happen before the return is filed, not after.

PwC estimates that roughly half of potential refunds relate to intercompany transactions.²⁸ If your U.S. entity is a limited-risk distributor, whether the refund has to be pushed to the foreign principal is a live transfer pricing question with documentation consequences.

If You Are Selling the Business — or Borrowing Against It

Two places the refund shows up that owners rarely anticipate.

In a sale, nobody has agreed who owns the claim. Standard purchase agreements were not written with a Supreme Court tariff reversal in mind, and most contain no language allocating refund entitlements at all. Sellers argue they bore the cost through a depressed valuation. Buyers argue the contingent claim came along with the assets. Both positions are reasonable, which is why this is now a real diligence issue.²⁹

Understand the asymmetry, because these are two different adjustments and they do not offset. Removing the 2025 tariff cost from historical EBITDA raises the earnings base — and at a multiple, that raises enterprise value. Treating the 2026 refund as non-recurring removes it from the base entirely — so the cash accrues to whoever contractually owns it, not to the multiple.

A seller generally wants both: the cost normalized out of history, and the refund cash allocated to itself. A buyer wants the mirror image. Quality-of-earnings providers are currently crediting only realized or highly probable refunds and treating the rest as contingent.

The sleeper risk is administrative. If a deal closes in late 2026 and the agreement doesn't say who manages the claim — who files, who decides whether to sue before February 4, 2027, who cooperates on records — a seven-figure claim can lapse through nothing but inattention. Cooperation covenants on this may need to run years past closing.

In a credit agreement, the refund may help less than it looks. Because it lands in COGS, it inflates net income and therefore EBITDA, which appears to create covenant headroom. But many agreements exclude “extraordinary, unusual, or non-recurring items” from consolidated net income, and a lender has an obvious argument that a one-time tariff refund is exactly that.³⁰ Read the EBITDA definition before you rely on the headroom.

The Fiscal Picture, Stated Honestly

Total public debt crossed $40 trillion for the first time on August 18, 2026 — $40.05 trillion, of which about $32.3 trillion is held by the public and the rest by government trust funds.⁶ It has since oscillated around the line.

The flow behind the stock: through July, the FY2026 deficit is $1.799 trillion, already larger than the entire FY2025 deficit of $1.775 trillion with two months still to run. CBO now projects about $2.1 trillion for the full year.³¹

Customs duties ran net negative $8.5 billion in July, because refunds of $33.4 billion exceeded collections of $24.8 billion. That was the third consecutive month of net outflow. June was the record: $49.2 billion refunded against $23.6 billion collected. Through all of 2025, monthly customs refunds never once exceeded $2 billion.³²

Which is the cleanest possible demonstration of a point the debate usually has to argue: tariffs cannot close the deficit. Customs duties are 3.4% of federal receipts. The deficit is $1.8 trillion and customs duties are $154 billion. You would need to multiply tariff revenue roughly twelve-fold to close the gap — before accounting for the demand destruction and retaliation a tariff of that size would cause. Individual income taxes alone are fifteen times customs duties.³¹

The spending side explains why. Social Security, Medicare, net interest, health including Medicaid, and defense are 78% of everything the federal government spends. Add income security and veterans and you are at 94%. And net interest has now passed defense — $931 billion against $804 billion in the first ten months of FY2026, the third-largest line in the budget, and roughly one of every five dollars the government collects.³¹

The mechanism that makes this durable is worth understanding, because it is not about future rate moves. The average interest rate on total marketable debt is 3.44%.³³ The 10-year Treasury is 4.69% and the 30-year is 5.23%. Every maturing low-coupon bond rolls into a higher-coupon one. Interest expense keeps rising for years even if yields stop rising today.

Treasury's response has been at the margins. On August 19 it announced it would double long-end buyback operations — from a $2 billion maximum to a $4 billion minimum per operation — beginning September 9, describing the purpose as liquidity support rather than rate management. Yields fell about nine basis points on the news and gave most of it back within two days.³⁴

What This Does, and Does Not, Mean for a Portfolio

Three things, and the order matters.

First, today's rates are not historically extreme. The 10-year Treasury closed at 4.69% on August 20. Compare that to the same month in prior years and the picture is unremarkable: today's level is below every January reading from 1968 through 2002 — a thirty-five-year stretch — and above every January reading since 2003 with one exception, January 2007 at 4.76%. The record high was 15.32% in September 1981. The record low was 0.62% in July 2020, and January 2021 printed 1.08%.³⁵ What feels like a crisis is a return to something near the long-run norm from a genuine anomaly, and the practical consequence for a taxable investor is that bonds and municipals pay something again.

Second, the evidence on debt and equity returns is more boring than the headlines. In a study covering more than twenty countries from 1975 to 2022, there were 153 country-years in which government debt exceeded 100% of GDP; stocks were positive in 104 of them, roughly two-thirds — in line with base rates generally. Italy and Belgium each spent more than 30 of 48 years above 100% debt-to-GDP and averaged 10.8% and 12.0% annual equity returns. Japan has been above 200% since 2010. U.S. debt-to-GDP roughly doubled from 2008 to today while the S&P 500 compounded in double digits. Debt levels are slow-moving and universally known, which is exactly the sort of information markets have already priced.³⁶

The honest counter-argument, which deserves airing: some researchers contend that deficits inflate corporate profits through sectoral balances — meaning today's record margins are partly deficit-financed and would compress under fiscal consolidation.³⁷ That is a different mechanism and it is not obviously wrong.

Third, the deficit is widest when equities are cheapest, and that is arithmetic rather than coincidence. Recessions collapse tax receipts and trigger automatic stabilizers and stimulus simultaneously. The FY2009 deficit was the largest to that point; the S&P 500 bottomed in March 2009. The FY2020 deficit of $3.1 trillion is the all-time record; the S&P 500 bottomed in March 2020. An investor who sold on “the deficit is exploding” in either year sold the bottom.³⁸

Since 1970 the federal government has run a deficit in every year but four — 1998 through 2001 — while a balanced portfolio compounded through all of it.³⁸

On the midterms, in November: midterm years have historically been below average for equities, not above. Across the fourteen midterm years since 1970, the S&P 500's average total return is roughly 3.6%, against about 12% for all years in the period — with 2022 at −18.1% and 2018 at −4.4%. Post-election six-month returns have been stronger, in the low-to-mid teens depending on the sample and start date.³⁹ All of it rests on fourteen observations, which is not enough to forecast anything, and I would not build a position on any of it.

Questions Worth Asking

If you import anything, or if you're planning a transaction in the next two years:

1. 1

   Whose importer-of-record number is on our entries, and have we filed?

   If the answer is a freight forwarder or a supplier, the refund is going to them. Ask what they have filed and what they intend to remit.

2. 2

   Are we enrolled in ACH refund in the ACE portal, with current banking details?

   This is a five-minute check that is currently the sole reason $1.6 billion has not been delivered.

3. 3

   Do we have finally liquidated entries, and what is our plan before February 4, 2027?

   The answer is file, sue, or knowingly accept the risk. There is no fourth option.

4. 4

   Did we itemize tariff surcharges to customers — and did anyone itemize them to us?

   One creates a contingent liability against the refund. The other may create a claim.

5. 5

   What does the refund do to our 2026 tax, and do our estimates already reflect it?

   Specifically: was the duty expensed or capitalized, is the inventory sold, and are we a large corporation that has lost the prior-year safe harbor?

6. 6

   If we are selling, who owns the pending refund claim under the LOI as drafted?

   If the agreement is silent, that is a negotiation you are about to have anyway — better before signing.

If you own a business working through any of this, we're glad to talk it through — including if what you need is a customs or tax specialist rather than an advisor.

The Bottom Line

The refund is a filing deadline, a tax event, and a contract question long before it is a market event — and the last of the three is the only one the financial press is covering.

The $40 trillion number is real, the trajectory is genuinely difficult, and the arithmetic says tariffs were never going to fix it. It is also true that debt levels have been a poor guide to portfolio decisions for fifty years, and that deficits look worst precisely when markets look best in hindsight.

Both things can be true. The difference is that one of them has a February deadline.