US Customs and Border Protection has postponed the third and final phase of the refund tool it built to return an estimated $166 billion in tariffs that the Supreme Court struck down in February 2026. Phase 3 of the Consolidated Administration and Processing of Entries system, known as CAPE, had been scheduled to open on August 20. It is now delayed with no replacement date, according to trade counsel tracking the rollout and to CBP’s own filings with the Court of International Trade.
The delay matters because Phase 3 is the tranche that covers “finally liquidated” entries: roughly $11.4 billion in duties on shipments where CBP already closed the books and the ordinary administrative window has shut. That is about 6.9% of the total refund pool, and it is the hardest 6.9% to pay.
For retailers and e-commerce importers, this is the plumbing behind a line item that has been flattering quarterly results all summer. Refund cash has already covered price investments at off-price chains and padded margins at apparel groups. Whether the remainder arrives, and who is legally entitled to it, is now a live question rather than a scheduling detail.
In short
- CAPE Phase 3 is postponed indefinitely. It was set for August 20, 2026 and has no new launch date, per an update circulated on August 25.
- About $11.4 billion sits in the delayed tranche, covering finally liquidated entries, roughly 6.9% of the $166 billion pool.
- Payments so far are substantial but incomplete. As of August 21, 2026, about $132.5 billion had entered CAPE and roughly $106.6 billion was certified to Treasury.
- Around $1.7 billion of approved refunds cannot move because importers have no valid ACH banking details on file, an issue that hits non-resident sellers hardest.
- The legal split is the real risk. Importers who filed at the Court of International Trade are covered by reliquidation orders; those who did not may recover nothing on finally liquidated entries if the government wins its appeal.
What CAPE Phase 3 covers and why it stalled
CAPE is the electronic mechanism CBP built inside the Automated Commercial Environment to process refunds at scale after the February 2026 ruling. Rather than force roughly 330,000 importers to file individual claims across more than 53 million entries, CBP consolidated the work into declarations that sweep many entries at once. The agency rolled it out in stages, each stage defined by how far along the customs lifecycle an entry had travelled.
Phase 1 opened on April 20, 2026 and handled the simplest population: entries still unliquidated, plus entries liquidated within 80 days of the submission date. Phase 2 followed on June 29, 2026 and extended coverage to entries flagged for Reconciliation where no reconciliation entry had been filed. Both phases dealt with entries where CBP retained clear administrative authority to reopen and adjust.
Phase 3 is different in kind, not just in sequence. It targets entries that have been finally liquidated, meaning CBP completed the duty calculation and the statutory window to reopen it administratively has closed. Recovering money on those entries requires reliquidation, which is a more intrusive act than simply adjusting an open file.
The reason CBP gave for the delay
According to CBP’s representations to the Court of International Trade, the agency temporarily delayed deployment in order to add validations. The stated concern is technical rather than political: reliquidating an entry to strip out unlawful tariffs risks inadvertently altering the non-IEEPA duties sitting on the same entry.
That is a real problem at scale. A single apparel or footwear entry can carry ordinary most-favored-nation duty, Section 301 duty, Section 232 duty, and antidumping or countervailing deposits alongside the struck-down tariffs. A crude reliquidation that recalculates the whole entry could hand back money the importer was never owed, or claw back money it legitimately paid.
Trade counsel publishing on the postponement, including Diaz Trade Law, reported the delay on August 21 without an accompanying explanation from CBP at that time. The validation rationale surfaced later through the court docket. Neither the agency nor the reporting firms have offered a revised target date.
Why “finally liquidated” is the expensive category
Liquidation is the point at which a customs entry becomes final. Under normal practice an importer has 180 days from liquidation to file a protest, and once that window closes the entry is ordinarily immune from reopening. The IEEPA refund population includes a large tail of entries that liquidated long before the Supreme Court ruled.
Those entries are not defective in any way the importer caused. They simply aged out of the administrative process while the litigation was pending. That timing accident is now the difference between an automatic refund and a lawsuit.
How much of the $166 billion has actually been paid
The headline progress is genuine. CBP has moved the majority of the pool through the system in roughly four months, which is fast by the standards of customs administration. The remainder, however, is concentrated in the categories that are structurally hard to clear.
The table below tracks the disclosed milestones. Figures come from CBP filings and from trade advisers monitoring the docket, and they measure slightly different things: money “accepted for processing” is not the same as money certified to Treasury, which is not the same as money in an importer’s account.
| Milestone date | Phase status | Accepted into CAPE | Certified to Treasury | Share of $166bn pool certified |
|---|---|---|---|---|
| April 20, 2026 | Phase 1 opens | Rollout begins | Not yet material | Under 1% |
| June 29, 2026 | Phase 2 opens | ~$121.75bn | ~$86.3bn | ~52% |
| August 4, 2026 | Phase 2 running | Not separately disclosed | ~$100bn | ~60% |
| August 21, 2026 | Phase 3 pending | ~$132.5bn | ~$106.6bn | ~64% |
The operational counts tell a similar story of bulk progress with a stubborn tail. CBP has recorded 272,029 CAPE declarations submitted, of which 191,494 passed file validations. Some 26.4 million entries have been accepted for removal of the struck-down duties, and 18.76 million entries have been liquidated or reliquidated.
The gap between 272,029 declarations submitted and 191,494 passing validation is worth noting. Roughly 30% of submissions have failed validation at least once, which usually means data quality problems rather than ineligibility. Those filings are recoverable, but each one is a delay measured in weeks.
Why the pace slowed after June
The month-by-month shape of the disbursements is informative. Between the Phase 2 opening on June 29 and August 4, certified amounts climbed from roughly $86.3 billion to about $100 billion. Between August 4 and August 21, they added only a further $6.6 billion or so.
That deceleration is not a sign of administrative failure. It reflects the fact that the easy population was cleared first, and what remains requires either a software capability that does not yet exist or a legal determination that has not yet been made.
Importers modelling refund receipts on the June to July run rate will therefore overshoot. The realistic planning assumption is that Phase 1 and Phase 2 entitlements continue to arrive on the normal cycle while Phase 3 entitlements sit unresolved into the fourth quarter.
What the outstanding balance actually consists of
Approximately $59.4 billion of the pool remained unpaid as of August 21, or about 36%. That figure is not a single blockage. It breaks into at least four distinct buckets, each with a different fix and a different probability of ever being paid.
- In-flight processing: declarations accepted but still inside the normal 60–90 day window between CAPE acceptance and disbursement.
- Validation failures: submissions rejected on data quality that importers can correct and resubmit.
- Banking blockages: approved refunds with no valid ACH destination, discussed in the next section.
- The Phase 3 tranche: roughly $11.4 billion on finally liquidated entries, contingent on both the software and the appeal.
Only the last bucket carries genuine legal risk of non-payment. The first three are friction, and friction resolves. That distinction matters for anyone modelling refund cash into a forecast.
Why $1.7 billion of approved refunds cannot leave Treasury
One of the more avoidable failures in the programme has nothing to do with tariffs, courts, or eligibility. CBP issues all IEEPA refunds electronically by Automated Clearing House transfer, and valid banking information must be on file in ACE before the money can move. Where it is not, the refund is approved and then simply sits.
As of August 21, about $1.7 billion covering 22,170 refunds was pending ACH enrollment. A related tally puts roughly 9,837 refunds as untransmitted specifically because the importer had not supplied account information. The two counts measure overlapping but distinct stages, and neither is a rejection: the money is allocated and unclaimed.
This is the same category of defect that is about to cost importers their ability to file entries at all. CBP’s parallel enforcement push means importer of record numbers can be voided from September 18 where Form 5106 data is inaccurate or incomplete. Stale contact and banking records now create two separate failures from one root cause.
The non-resident importer problem
The banking requirement lands hardest on non-resident importers, a population that includes a large share of cross-border marketplace sellers. An ACH transfer requires a US bank account. An overseas seller that imported into the United States, paid the duties, and filed a valid refund claim may still have no practical way to receive the money.
These are not marginal filers. Non-resident importers of record became far more common as marketplaces pushed sellers toward delivered-duty-paid arrangements, and many of them ran genuine import programmes without ever opening US banking relationships. The refund mechanism assumes a domestic corporate treasury that a meaningful slice of the claimant pool does not have.
The Form 4811 workaround and its limits
The practical route for importers who genuinely cannot enroll in ACH is to designate a third party to receive the refund. That is done by filing Form 4811, the Special Address Notification, naming a US licensed customs broker as the recipient. The broker must in turn have valid ACH details on file.
The workaround is real but it is not free. It puts a large sum of the importer’s money into a broker’s account, which raises questions about segregation, timing of onward transfer, and what happens if the broker relationship ends mid-process. Importers using this route should paper it properly rather than relying on an existing power of attorney.
It also does not solve the underlying record-keeping problem. An importer that cannot receive a refund because its ACE data is stale is likely to have the same stale data on the entry filings that generated the refund in the first place.
Who gets paid automatically and who has to sue
The Phase 3 tranche is where administrative process gives way to litigation, and the split is unusually clean. It depends almost entirely on whether an importer filed a protective action at the Court of International Trade before its entries went final.
On July 15, 2026 the CIT signalled that it would issue case-specific orders across roughly 3,700 individual IEEPA cases, each directing CBP to reliquidate the named plaintiff’s entries. A subsequent order in mid-July directed CBP to reliquidate, without regard to the struck-down duties, any of those plaintiffs’ entries that had been liquidated for more than 80 days. The relief reaches entries already final and entries becoming final during CAPE processing.
Critically, that relief is individualized. It runs to the plaintiffs in those roughly 3,700 cases and not to the wider population of importers who paid the same unlawful duties on the same kinds of goods.
The government’s appeal
The Administration has appealed to the Federal Circuit, challenging whether the CIT has authority to grant relief beyond the named plaintiffs. The Department of Justice position is that CBP lacks statutory authority to reliquidate finally liquidated entries unilaterally for non-litigants, given the finality that attaches under the customs statute.
If the government prevails, the practical outcome is stark. Importers with pending CIT cases receive reliquidation on their finally liquidated entries, and importers without suits may be unable to recover on theirs at all.
That appeal is running alongside a separate and equally consequential trade case. Retail importers are already watching a September 15 court test on forced-labor tariffs where refunds are also at stake, which means two distinct refund questions are moving through the courts in the same window.
Remaining routes for non-litigants
Importers who did not file are not automatically out of options, but every remaining route is worse than having sued early. The 180-day protest window runs from each entry’s individual liquidation date, so for older entries it has already closed.
| Importer position | Route to refund | Timing risk | Dependent on appeal? |
|---|---|---|---|
| Filed protective CIT action | Court-ordered reliquidation via Phase 3 | Waits on CAPE Phase 3 software | No |
| Entry liquidated under 180 days ago | File a protest under the standard window | Window closes on a rolling basis | Partially |
| Entry liquidated over 180 days ago, no suit | New CIT suit or await class treatment | High | Yes |
| Entry unliquidated or recently liquidated | CAPE Phase 1 or 2, already open | Low, normal 60–90 day cycle | No |
The asymmetry in that table is the argument for acting now. Filing costs are modest relative to the sums at stake, and the downside of waiting for a favourable appellate ruling is losing the ability to claim at all.
What the refunds have already done to retail results
The refund pool stopped being an abstraction for retail investors around the second-quarter reporting season. Several chains have disclosed material refund receipts, and in at least one case the money changed commercial strategy rather than simply landing in profit.
Burlington disclosed a $55 million refund and put the entire amount into price investment rather than banking it, a decision that helped comparable sales while pressuring the third-quarter guide. The company’s choice to spend the tariff refund on lower prices is the clearest example of refund cash flowing straight to the shelf edge.
Dillard’s took a different route, booking a $37.2 million refund that lifted quarterly profit by 34%. PVH Corp is expected to carry roughly $100 million of refund benefit through its second-quarter report. The pattern is consistent: the money is large enough to move a quarter, and small enough that it is easy to miss in a headline number.
| Retailer | Disclosed or expected refund | How it was used or reported | Reporting context |
|---|---|---|---|
| Burlington Stores | ~$55m | Fully reinvested in price cuts | Q2 FY2026 results |
| PVH Corp | ~$100m | Expected to carry the quarter | Q2 FY2026, reporting September 2 |
| Dillard’s | $37.2m | Contributed to a 34% profit rise | Q2 FY2026 results |
| Five Below | Not separately quantified | Guidance framed around lapsed tariff assumptions | Q2 FY2026, reporting September 2 |
| Lululemon | Offsets a ~$380m tariff bill | Guidance pressure rather than benefit | Q2 FY2026, reporting September 3 |
Analysts covering the PVH second-quarter report due September 2 will be pressing on exactly this point. The useful question is not how large the refund was, but which CAPE phase produced it, because that determines whether the next tranche is a formality or a lawsuit.
Why the phase matters more than the amount
A refund received through Phase 1 or Phase 2 tells you very little about what remains. Those phases cleared the administratively easy population, and a company that received money there may still have significant exposure sitting in finally liquidated entries.
Conversely, an importer with no Phase 3 exposure has effectively been paid in full and should not be modelled as having further upside. The disclosure quality across retail has not yet caught up to this distinction, and most filings describe refunds as a single undifferentiated number.
How the money shows up in the accounts
Treatment varies, which complicates comparison across the sector. Where the original tariff was capitalised into inventory cost, the refund tends to flow back through cost of goods sold as that inventory sells, spreading the benefit across quarters. Where it was expensed, the refund is more likely to appear as a discrete item.
Interest is an additional wrinkle. CBP’s certifications to Treasury include interest, which for duties paid across 2025 and into 2026 is not trivial on large balances. That interest component is generally reported below the gross margin line even where the principal is not.
The practical effect is that two retailers with identical refund entitlements can show very different margin optics in the same quarter. Reading the cash flow statement alongside the margin bridge is the only reliable way to separate refund cash from underlying trading performance.
The reversal risk nobody is provisioning for
There is a scenario, currently unquantified in most disclosures, in which money already received is exposed to the appellate outcome. Refunds paid under Phase 1 and Phase 2 rest on CBP’s own administrative authority over open entries and look secure. Amounts paid on reliquidated finally liquidated entries rest on court orders that the government is actively appealing.
No major retailer has disclosed a provision against refund reversal, and the reporting to date gives no indication that CBP is seeking to recover disbursed amounts. Importers with large Phase 3 entitlements should nonetheless be tracking the Federal Circuit docket rather than treating the cash as unconditionally theirs.
What importers should do before Phase 3 opens
The delay is an unusual gift: it creates time to fix the things that will otherwise stop a refund on the day the tool goes live. Most of the work is data hygiene rather than legal strategy, and it overlaps almost exactly with the compliance work already required by the September enforcement changes.
- Pull current Form 5106 data from the ACE Secure Data Portal and confirm the physical address, phone number, and email all belong to the importer rather than a broker or forwarder.
- Verify ACH enrollment in ACE before any Phase 3 submission, since an approved refund with no banking destination simply waits.
- Inventory entries by liquidation status, separating unliquidated, recently liquidated, and finally liquidated populations, because each maps to a different phase and a different legal posture.
- Check the 180-day protest clock on every entry that liquidated recently, as this window closes on a rolling per-entry basis.
- Assess whether a protective CIT filing is warranted for material finally liquidated exposure, given that non-litigants may recover nothing if the government wins.
- Rework failed CAPE declarations rather than resubmitting them unchanged, since roughly 30% of submissions have failed validation at least once.
- Confirm power of attorney is executed directly with the importer where a broker is filing or receiving on its behalf.
For non-resident importers the sequencing is different and more urgent. Opening a US banking relationship takes longer than any of the other steps, so that decision should be made now rather than when Phase 3 finally deploys.
How the delay collides with the September compliance calendar
The postponement does not arrive in a quiet month. September brings a cluster of customs changes that draw on the same internal resources importers would otherwise point at refund recovery, and in several cases on the same underlying data.
From September 18 CBP can void importer of record numbers where Form 5106 data is inaccurate or incomplete, an enforcement step flowing from the June 2026 executive order on customs enforcement. On September 22 the agency begins its Entry Type 13 test for electronic informal entry of international mail, the process built to replace the suspended de minimis exemption. Both changes demand accurate importer records.
That overlap is an argument for treating the work as one project rather than three. The physical address, phone number, email, and ACH details that keep an importer of record valid are the same fields that determine whether a refund can be disbursed.
Resourcing the work at the wrong time of year
The timing is awkward for retail specifically. September and October are when import volumes peak ahead of the holiday season, which is precisely when customs teams have least capacity for remediation projects.
Companies that defer the data work until Phase 3 opens will be doing it during peak. Those that use the delay to complete it will be positioned to file on day one, which matters if the tranche is processed in submission order.
The disclosure question for the next reporting cycle
Investors have so far accepted refund figures as single numbers without much scrutiny of their composition. That is unlikely to survive another quarter in which some retailers collect and others wait.
The disclosure that would actually be useful is a split between amounts received, amounts entitled and pending under Phases 1 and 2, and amounts contingent on Phase 3 and the appeal. Very few filings currently offer it, and the ones that do will be easier to underwrite.
What to watch next
Three dockets and one piece of software determine how the remaining $59.4 billion resolves. None of them has a firm public date, which is itself the most useful thing to understand about the current position.
The Federal Circuit appeal on universal relief is the largest single variable, because it decides whether roughly $11.4 billion reaches non-litigants or only the plaintiffs in about 3,700 cases. CBP’s Phase 3 deployment is the second, and the agency has committed to validations without committing to a schedule.
The third is the steady administrative grind of validation failures and banking enrollments, which will keep releasing money regardless of how the appeal lands. That grind is where importers have direct control, and it is the reason the practical advice is unglamorous.
The wider tariff calendar has not paused for any of this. New duties keep taking effect while refunds for struck-down duties are still being processed, and importers are managing both flows simultaneously. Anyone reconciling a refund position this autumn is doing it against a moving duty base rather than a settled one.
Frequently asked questions
What is CAPE and why did CBP build it?
CAPE stands for Consolidated Administration and Processing of Entries. CBP built it inside the Automated Commercial Environment to process refunds at scale after the Supreme Court held certain IEEPA tariffs unlawful in February 2026, rather than requiring roughly 330,000 importers to claim individually across more than 53 million entries.
Why was CAPE Phase 3 delayed?
Phase 3 was scheduled for August 20, 2026 and has been postponed with no new date. According to CBP’s representations to the Court of International Trade, the agency needs to add validations so that reliquidating an entry to remove the struck-down duties does not inadvertently alter the non-IEEPA duties on the same entry.
How much money is still outstanding?
Roughly $59.4 billion of the estimated $166 billion pool remained unpaid as of August 21, 2026, or about 36%. That includes in-flight processing, failed validations, refunds blocked on missing banking details, and the roughly $11.4 billion Phase 3 tranche on finally liquidated entries.
Do I need to file a lawsuit to get my refund?
Not for entries that are unliquidated or recently liquidated, which are handled administratively through Phase 1 and Phase 2. For finally liquidated entries, importers who filed protective actions at the Court of International Trade are covered by reliquidation orders, while non-litigants may be unable to recover if the government wins its Federal Circuit appeal.
What is the 180-day protest window?
It is the standard period during which an importer can protest a customs liquidation, running from each entry’s individual liquidation date. Because it is per-entry and rolling, it has already closed on older entries while remaining open on more recent ones, so the position must be assessed entry by entry.
Why can’t some importers receive their refunds?
CBP issues IEEPA refunds electronically by ACH transfer, which requires valid banking information on file in ACE. About $1.7 billion covering 22,170 refunds was pending ACH enrollment as of August 21, 2026, and the problem falls hardest on non-resident importers who have no US bank account.
What can a non-resident importer do without a US bank account?
The practical route is to file Form 4811, the Special Address Notification, designating a US licensed customs broker as the refund recipient, provided that broker has valid ACH details on file. This should be documented specifically rather than assumed under an existing power of attorney, given the sums involved.
How long does a refund take once CAPE accepts a declaration?
Valid refunds are generally issued within 60–90 days of acceptance of the CAPE declaration, unless a compliance concern triggers further CBP review. Certifications to Treasury include interest, which is material on balances accumulated across 2025 and 2026.
Could refunds already received be clawed back?
There is no reporting to suggest CBP is seeking to recover disbursed amounts, and refunds paid under Phases 1 and 2 rest on CBP’s own authority over open entries. Amounts paid on reliquidated finally liquidated entries rest on court orders that the government is appealing, so importers with large Phase 3 entitlements should track that docket rather than treat the cash as unconditional.