Section 338 and the 50% Canada Tariff: The Procurement Playbook, Not the Legal Explainer
Three presidential proclamations signed 20 July 2026 impose an additional 50% duty on roughly $17.7 to $20 billion of Canadian goods, and USMCA origin does not exempt a single covered line. This is the first use of Section 338 of the Tariff Act of 1930 in the statute’s history. The duty was scheduled to attach at 12:01 a.m. Eastern on 19 August 2026, was paused at the deadline while the two governments talked, and took effect at 12:01 a.m. Eastern on 22 August 2026 when that pause expired with no deal signed.
For procurement teams the operative fact was never the rate. It is that the free-trade-agreement shield every North American sourcing decision of the last six years was built on failed to apply, and that the duty attaches on entry for consumption, which means goods already sitting in a US bonded warehouse are not safe. This particular 50% is now being collected, and the mechanism behind it is proven and sitting on the shelf for whoever reaches for it next. That is the part worth learning from, and it is why this piece is written as a case study rather than a countdown.
Status as of 23 August 2026. The talks collapsed and no deal was reached before the pause expired. The 50% Section 338 duty took effect at 12:01 a.m. Eastern on 22 August 2026 as proclaimed, and CBP issued implementing guidance on 21 August (CSMS #69606660) confirming the effective date and the HTSUS reporting headings importers must file under. The duty is being collected on covered entries for consumption dated 22 August or later. Canada has responded: Prime Minister Carney suspended negotiations on 21 August and announced dollar-for-dollar retaliatory tariffs from 8 September 2026 on steel, dairy, appliances, agricultural equipment, pulp and paper, and electronics. No suspension, withdrawal or court challenge to the US duty had been filed as of this publication. Check CBP’s CSMS messages for any subsequent amendment before acting on any figure in this piece.
What is Section 338 of the Tariff Act of 1930, and why does it matter more than the rate?
Section 338 lets the President declare that a foreign country has discriminated against US commerce and impose new or additional duties by proclamation alone. There is no prior investigation by the International Trade Commission, no USTR docket, no Commerce national-security finding, which is what distinguishes it from Section 301 and Section 232. Holland & Knight’s read is that the statute caps the duty at 50% ad valorem but permits escalation to a full import ban if the discrimination continues, and that the President may suspend, amend or revoke at any time.
Read those two facts together and you have the planning problem in one line. The rate is already at its statutory ceiling, so there is no room for a worse number, but there is room for a worse instrument. And unlike the Section 122 surcharge that ran on a 150-day clock, Section 338 carries no expiry, so nothing removes it except a decision.
The stated rationale, per Ambassador Greer’s statement, is three specific Canadian measures: US alcohol pulled from provincial shelves, better dairy market access granted to the EU than to the US, and a cap on US vehicle exports from reshoring companies. The White House fact sheet puts numbers on it, including US vehicle exports to Canada down 22% ($5.6 billion) year over year and alcohol down 81% ($582 million).
Does USMCA protect Canadian-origin goods from the Section 338 tariff?
No. The proclamations apply “to all covered goods regardless of whether a good originates under the U.S.-Mexico-Canada Agreement,” and the duty stacks on top of any other duties, taxes and fees already owed rather than replacing them. Troutman Pepper Locke put it bluntly: none of the three proclamations creates a USMCA carve-out.
The timing is the part worth sitting with. The USMCA Free Trade Commission held its joint review on 1 July 2026, and White & Case reports that the United States declined to extend the agreement, which triggers annual reviews under Article 34.7.4 running until either an extension is agreed or the agreement expires on 1 July 2036. The agreement stays in force. Preferential tariffs, rules of origin and dispute settlement all remain operative.
So the agreement was not broken. It was routed around. That distinction matters for planning, because it means a Canadian lane can be fully USMCA-compliant, fully documented, and fully exposed, at the same time, and your certificate of origin file tells you nothing about your risk.
Which Canadian products are actually covered by the 50% Section 338 tariff?
This is where every product-name-based exposure screen breaks, and it is the single most useful thing on this page.
The three proclamations are named for motor vehicles, alcoholic beverages and dairy. The annexes do not match the names. Global Trade Alert’s line-level count found that although motor-vehicle goods account for 19 of every 20 dollars in scope, none of the 439 traded lines sits in the tariff schedule’s vehicles chapter, with the closest items being car seat parts classified as furniture. Vehicles and auto parts are excluded because they already carry Section 232 duties, and the Section 338 proclamations exclude anything already subject to Section 232.
| Proclamation (signed 20 Jul 2026) | What the name implies | What the annex actually covers | USMCA shield | Stacks on existing duties | Practical screening failure this causes |
|---|---|---|---|---|---|
| Motor vehicles (FR 2026-14997) | Cars, trucks, auto parts | No vehicles-chapter lines at all. Honey, plants, wood, lumber and plywood, textiles and apparel, tools, cosmetics, chemicals, plastics, machinery, toys, furniture, sporting goods, fine art | None | Yes | Auto buyers relax, and the furniture, plywood and textile buyers who are actually exposed never look |
| Alcoholic beverages (Proc. 11046, FR 2026-14991) | Beer, wine, spirits | Beer, wine and spirits, plus grapefruit essential oil, wooden tableware, kraft and greaseproof paper, coated paperboard, and ice and field hockey equipment | None | Yes | A packaging buyer sourcing Canadian coated paperboard has no reason to open an alcohol proclamation |
| Dairy (FR 2026-14992) | Cheese, milk | Milk and cream, whey and protein concentrates, lactose, casein and derivatives | None | Yes | Food and nutraceutical formulators buying whey or casein as an ingredient, not as “dairy”, miss it entirely |
| Carved out across all three | Energy, potash, fish, critical minerals, goods already under Section 232 (steel, aluminium, copper, autos and auto parts), WTO Civil Aircraft Agreement goods | n/a | n/a | Assuming a carve-out applies without checking the eight-digit line |
The instruction that falls out of this table is not subtle. Screen at the eight-digit HTS line against the annexes, per SKU, not by product family and not by proclamation title. If your exposure analysis was built by asking category managers “do you buy anything Canadian in autos, alcohol or dairy,” it is wrong, and it is wrong in the direction that costs money.
Why does a single tariff line reset a sourcing strategy?
Tariff and non-tariff barriers are the biggest levers stopping supply chains from running autonomously, and this is exactly that situation playing out again. We have watched the pattern repeat for a decade. The Qatar embargo forced corporations across a deeply interdependent GCC to scramble for new supply routes. The on-again, off-again US and China trade war ended with perishable commodities being dumped at sea. Panama, Suez, Hormuz and Bab el-Mandab each turned into a freight choke point in their own turn, and every one of them forced global supply chains to pivot faster than they were built to, because the cost differential on goods in transit is enough to put a P&L into the red on its own.
Which lands you back at the two oldest questions in the trade, make versus buy and diversify versus centralize. Under the new US duty regime, corporates that kept a manufacturing footprint in the US started this with a head start on their competitors, which is exactly what happened the last several times a caveat like this appeared.
Where does the duty attach, and why is the warehouse reflex wrong this time?
The duty applies to goods entered for consumption, or withdrawn from warehouse for consumption, on or after the effective date, which after the pause expired is 22 August 2026. The two consequences below are the ones that catch teams out, and both invert the usual instinct.
Bonded inventory is not a shelter. Per Troutman’s analysis, duty attaches on withdrawal for consumption, not on arrival or on warehousing, so stock already sitting in a US bonded warehouse and withdrawn on or after the effective date pays the full 50%. Parking it does not save it. It only defers the bill and adds carrying cost.
Foreign trade zones lock the rate in, not out. Covered merchandise admitted to a US FTZ on or after the effective date must generally be admitted under privileged foreign status, which fixes classification and duty rate as of the admission date. C.H. Robinson’s advisory gives the same instruction. An FTZ is a cash-flow tool here, not an exemption.
We have not found any importer exclusion or petition process for Section 338, which is consistent with a statute that requires no investigation to invoke. CBP issued implementing guidance on 21 August 2026, the day before the duty attached, confirming the effective date and the HTSUS reporting headings importers must file under. That guidance now supersedes the proclamation annexes as the operative reference for what you file. If you scoped your exposure off the annexes during the pause, rebuild it against CBP’s CSMS messages, and check for technical corrections before acting on any subheading in this piece.
Should you buy forward ahead of a tariff like this one?
Sometimes, and the honest answer depends on a probability nobody can give you, which is why most coverage skips the question.
Our tariff inventory playbook makes the general case: buffer inventory is a loan taken out against a guess, and the hedge only pays when the tariff lands in full and the stock turns on schedule. Section 338 changes two inputs in that model, in opposite directions.
Pushing toward buying forward: there is no expiry clock, so if it lands it stays until someone decides otherwise, and the duty is at the statutory ceiling, so the downside case is the full 50% rather than some fraction of it.
Pushing against, and this was our read while the clock was still running: the tariff is openly an opening bid. Holland & Knight read the 30-day runway as a window to extract concessions, USTR confirmed talks were continuing, and the President could suspend the proclamations at any point. A tariff designed as leverage in a live negotiation, we argued, carries a materially higher chance of being suspended than one designed as permanent policy.
Assume $1M of covered goods, four months of cover pulled forward, 20% annualised carrying cost (capital, space, shrink, obsolescence), so roughly $67K of carrying cost. The tariff avoided at 50% is $500K, which is a far more lopsided bet than the 10% case in our earlier piece.
| If the tariff… | Carrying cost | Duty avoided | Net | What it tells you |
|---|---|---|---|---|
| Lands and holds past your cover | $67K | $500K | +$433K | At a 50% rate the hedge is barely a decision on storable, non-perishable, non-obsolescing lines |
| Is paused or suspended before it attaches | $67K | $0 | -$67K | The loss is small relative to the win, which is the whole point |
| Lands, but stock sits 10 months instead of 4 | $167K | $500K | +$333K | Still positive. A 50% rate absorbs a lot of bad turns |
| Lands, and the goods are perishable or fashion-dated | $67K plus writedown | $500K | Depends entirely on the writedown | This is the only quadrant where the answer is clearly no |
That table is arithmetic, and arithmetic is the easy part. It leaves out three constraints that decide the question in practice and none of which appear in the model. Whether the supplier can physically ship four months of cover inside two weeks. Whether you have the working-capital facility to fund it without breaking a covenant. And whether pulling that hard on an allocation torches a relationship you will still need in 2027. The third is the one that gets left out of the spreadsheet and remembered at the next negotiation.
The first row of that table is the one this case turned on, and it is not the row we expected to be writing about. Before the deadline we argued that a tariff built as leverage carries a materially higher suspension probability than one built as durable policy. The reasoning held up. The conclusion did not. A pause did arrive hours before the 19 August clock ran out, which looked at the time like that argument playing out in public, and then it expired on 21 August with nothing signed and the duty attached the following morning. A team that bought four months of cover at speed is sitting on row one rather than row two: roughly $67K of carrying cost against $500K of duty it will not pay on that window.
We want to be precise about what that does and does not prove, because the temptation after a call like this one is to over-learn from it. Buy-forward has not become the default. Had the suspension held, the same team would be $67K down for nothing and this paragraph would read the other way round. What the outcome actually shows is the asymmetry in the table doing its job. The loss case was small enough that getting the direction wrong would have cost a fraction of what getting it right paid.
The reason the 50% rate flips the usual advice is scale, not certainty. At 10% the carrying cost eats the benefit. At 50% it does not, which means the question stops being “will it land” and becomes “can you physically and financially move the goods in time, and will they still be sellable.”
What does this do to the nearshore-to-Canada thesis?
It does not kill it, and anyone telling you it does is selling something. But it removes the assumption the thesis quietly rested on.
The nearshoring case for Canada and Mexico was never only about transit time. It was that a USMCA-qualifying lane was tariff-insulated, so you were buying policy stability along with the shorter lane. Section 338 is a demonstration that the shield can be routed around by a statute nobody had used in 96 years, with no investigation, no notice beyond 30 days, and no expiry. Global Trade Alert calculates that with the duty in effect, Canada’s trade-weighted average US tariff rises from 4.37% to 6.27%, and the share of Canadian exports USMCA can still shield falls from 85.6% in October 2025 to 82.3%.
That second number is the one to carry into your next sourcing review. The shield still covers most of the trade. It is just no longer a constant, and a sourcing model that treats FTA coverage as a fixed input rather than a policy variable with a decay rate is now measurably wrong.
One thing this does not settle, and we would rather flag it than paper over it. The obvious defensive move is a second source in Mexico, and it is not obviously risk reduction. The United States has already declined to extend USMCA and triggered the Article 34.7.4 annual reviews, so a Mexican alternate is exposed to the same policy variable on a similar clock. Diversifying inside the agreement is not the same as diversifying away from it, and a second source only counts as risk reduction when it sits under a different policy regime, not just a different flag.
The bite also varies by category more than a single article can responsibly generalise. A packaging buyer on coated paperboard, a nutraceutical formulator on casein and a furniture importer on plywood are all in scope through proclamations named for something else, and their substitution options are nothing alike.
Who owns this internally, procurement or trade compliance?
Trade compliance owns the classification and the entry mechanics. Procurement owns everything that costs money.
The failure mode we would expect is the handoff. Compliance produces a correct list of covered HTS lines, procurement receives it as a compliance document rather than a cost event, and nobody converts it into a supplier conversation until the invoices arrive in September. The Incoterms point is the sharpest version of this: a DDP seller of covered Canadian goods just inherited a 50% cost increase and will come back to renegotiate or fail, and that is a procurement problem discovered by reading a contract, not a customs form.
To be straight about the basis for that: it is reasoned from how the work divides, not from a survey of who wins the argument in practice. The tie-break worth watching is budget authority. Whoever owns the P&L line the duty lands on tends to end up owning the response, whatever the org chart says, which in most mid-market structures points at procurement holding the problem and compliance holding the data.
What to watch now
Canada’s retaliation, which is now the nearest dated event on the calendar. Prime Minister Carney suspended the negotiations on 21 August, calling the terms on offer uneconomic and recalling the negotiating team, and announced dollar-for-dollar retaliatory tariffs from 8 September 2026 covering steel, dairy, appliances, agricultural equipment, pulp and paper, and electronics. If you sell into Canada on any of those lines, that date belongs in your own exposure model, and this piece has so far been written from the import side only. CBP CSMS messages and any Federal Register technical corrections to the annexes, because the scope you screened against may move. Any formal withdrawal or amendment of the proclamations, which is the only thing that removes the authority, and any filing at the Court of International Trade, which is where a challenge would land if one comes. And H.R. 2464, the Repealing Outdated and Unilateral Tariff Authorities Act, which would repeal Section 338 outright, and which is worth tracking not because it is likely to pass quickly but because its progress is a decent read on how durable this instrument is.
One honest note on the numbers in this piece. USTR’s own estimate puts total exposure near $20 billion, about 5.2% of the $382 billion the US imported from Canada in 2025, while Global Trade Alert’s line-level count comes to $17.7 billion against a $364.9 billion 2025 base. The gap is a definitional one and neither is wrong. We flag it because a piece that quotes one number without acknowledging the other is telling you it did not look twice.
