The barrier moves faster than the building

On July 20 the White House signed three proclamations invoking Section 338 of the Tariff Act of 1930, imposing an additional 50 percent on a range of Canadian goods effective August 19. No president had used that authority before. Section 338 permits new or additional duties of up to 50 percent against countries found to discriminate against US commerce, and the proclamations cited three Canadian practices: provincial alcohol restrictions, dairy quota rules, and a surtax on US-made vehicles. The named categories were motor vehicles, alcoholic beverages and dairy, but the schedules reached considerably further: wine, cement, plywood, furniture, fishing rods, seeds, clothing, hockey sticks.

Late on August 18, hours before the duty was to attach, the president announced a three-day pause, writing that Canada and the United States, "subject to the finalization of documents, have a DEAL." Prime Minister Mark Carney put the postponement at the end of the day Friday and said substantial progress had been made, "although there is important work still to be done." A further proclamation moved the effective moment to 12:01 a.m. Eastern on August 22. Nothing was rescinded: the underlying proclamations remain in force, and the duty attaches tomorrow unless something replaces it today.

On February 20, in Learning Resources, Inc. v. Trump and Trump v. V.O.S. Selections, Inc., the Supreme Court held that the International Emergency Economic Powers Act does not authorize tariffs, invalidating both the reciprocal duties and the trafficking and immigration duties. The reasoning was textual before it was doctrinal: the statute never mentions tariffs, duties or taxes; a power to regulate importation is a power to control rather than to raise revenue; and no president had read the act as a tariff authority in nearly fifty years. Three justices added the major questions doctrine on top. Three dissented. In March the Court of International Trade confirmed the practical consequence: the duties were unlawful at the time they were imposed, not merely void going forward.

Customs and Border Protection built a refund module, CAPE, and opened it in two phases in April and June. As of July 31, more than 75,000 declarations had been submitted, 25.1 million entries had passed validation and 5.02 million had failed it, and approximately USD 128.68 billion in potential and certified refunds had been accepted for processing. None of it returns automatically. The importer of record or its broker has to claim it, and companies that miss the litigation windows in February and April 2027 simply do not get it. A duty that was paid for a year, planned around, priced into contracts and then unwound retroactively is a strange sort of input to build a plant against.

What the ruling removed was the across-the-board instrument. What remains is a collection of narrower authorities, each with its own scope, procedure and clock: Section 232 on national-security grounds, now covering steel, aluminum, copper, autos and pharmaceuticals; Section 301 for unfair trade practices, extended in July to forced-labor grounds; Section 122, capped at 15 percent and 150 days; Section 201 safeguards, which require an investigation first; and now Section 338. The headline rate is the least stable feature of any of them. The stable part is the machinery underneath: which product code claims you, which carve-out you qualify for, what origin you can prove.

The Section 338 schedule shows what that machinery does in practice. It covers about USD 17.7 billion of Canadian goods and lifts Canada's trade-weighted average US tariff from 4.37 to 6.27 percent. Motor-vehicle goods account for nineteen of every twenty dollars, though nothing from the vehicles chapter appears in the list. The single largest affected line, electric control boards at USD 1.73 billion, pays nothing. It is carved out under Section 232. Network equipment pays 5 percent by way of a civil-aircraft exemption. And USMCA, which covered 86 percent of eligible trade in 2025, provides no relief here at all: an origin certificate worth a great deal against one statute is worth nothing against another. Section 232 itself moved twice this year, in April onto the full customs value of covered articles and again in June, widening coverage while cutting certain industrial machinery to 15 percent through the end of 2027.

A factory is a fifteen-to-twenty-year commitment with two to four years of lead time before it produces anything. The barrier it is nominally responding to now moves on a schedule measured in days, in both directions, and can be voided after the fact by a court. Those horizons do not meet. The predictable consequence is that announcements outrun construction, and the aggregate data is consistent with exactly that: US manufacturing construction spending is down roughly 21 percent from its June 2024 peak of USD 239 billion, with electronics and semiconductors off 44 percent from their own 2024 high. Non-electronics categories grew 5.6 percent between February 2025 and March 2026, which after inflation is about 2.3 percent. Manufacturing employment sits around 1 percent below its April 2025 level. Of 227 public industrial firms that announced footprint changes in the second quarter of 2025, 87 said they were expanding in the United States. And the leading indicators still show, as the analysts who assembled that count put it, little evidence of a reshoring-driven boom. An earlier note argued that a tariff is not a factory. This is the sharper version of the same point: a tariff is not even a stable reason to build one.

The tempting reading is that this argues for waiting. It argues closer to the opposite. If the rate cannot be forecast, then the value in a capacity decision shifts away from the bet itself and toward the reversibility of the bet: how fast a qualified line can be stood up somewhere else, how much of the qualification work travels with it, and how well a firm can prove on demand where its value was added. That is the conclusion the thesis reaches from the manufacturing side: the durable layer is the framework (standards, data, certification, training, routing) rather than the machine sitting inside it. Trade policy arrives at the same place from the legal side. In this regime origin, classification and certification are not compliance overhead; they are the instruments that decide the number. A firm that can re-site a certified process in months and document its origin without launching a project has hedged something real. A firm that sited a plant against a particular rate has hedged a headline. Our dated, falsifiable bets sit in the predictions register, and this note scores none of them.

Four limits, and the first is large. The Canadian outcome is unresolved at the time of publication: the pause expires today, and a finalized agreement, a further extension, and the duty attaching on schedule are all still live. Every Section 338 figure above should be read as conditional. Second, the USD 128.68 billion is CBP's own processing figure, combining potential with certified refunds; it is not money disbursed, and the amount finally paid will be smaller. Third, the construction and employment series are national aggregates, and they cannot separate a project shelved for tariff uncertainty from one shelved for interest rates, a finished CHIPS-era build, or ordinary weak demand. The correlation here is not identification. Fourth, nothing above establishes that the sectoral statutes are stable, only that they are harder to void. Section 232 was modified twice in four months. Slower to change is not the same as predictable.

Load-bearing claims are checked against the primary document where one exists, and otherwise against at least two independent accounts. Litigation details come from law-firm analyses of the opinions; refund figures trace to CBP's own status declaration to the Court of International Trade; the tariff-schedule arithmetic is Global Trade Alert's, computed on 2025 customs values. Where a figure is an agency's self-reported processing total rather than a settled amount, the text says so.

White & Case: Trump administration imposes 50% tariffs on certain Canadian products in first use of Section 338 ·Holland & Knight, 50 percent opening bid: Canadian imports subject to Section 338 tariffs amid USMCA talks ·Wiley: President Trump imposes new 50% tariffs on certain Canadian imports ·Global Trade Alert: Section 338 tariffs raise Canada's average US tariff to 6.27% ·ABC News: Trump announces temporary pause on Canada tariffs, citing new deal ·CNBC: Trump pauses 50% scheduled tariffs on Canada for three days, announces deal with Ottawa ·CBC News: Trump pauses 50 per cent tariffs on Canadian goods for 3 days ·GHY International, Section 338 tariffs: how Canadian exporters should prepare ·Holland & Knight, Supreme Court strikes down IEEPA tariffs: what importers need to know now ·Stinson: Supreme Court invalidates IEEPA tariffs; recent developments accelerate the refund process ·BDO, IEEPA tariff refunds: key FAQs for importers ·Skadden: Tariff refund mechanism takes shape after the Supreme Court's IEEPA ruling ·White & Case: United States modifies steel, aluminum and copper Section 232 tariffs ·C.H. Robinson: Updates to Section 232 tariffs on steel, aluminum and copper ·IoT Analytics, US manufacturing reshoring boom: what the data says ·FRED, total construction spending: manufacturing in the United States