US imports didn't fall under tariffs. They changed address.

China supplied 17.4% of US goods imports in the year to September 2022. In the year to May 2026 it supplied 7.8%. Over the same period total US goods imports did not contract — they rose 4.8%, from $3,239.7bn to $3,394.6bn. The tariffs changed where American imports come from. They did not reduce how much America imports.

Every figure below uses trailing twelve-month sums rather than single months. The customs-basis country series are not seasonally adjusted, and a single-month comparison across them measures the calendar as much as the trade. All are monthly, in US dollars, and run 89 days behind: the most recent observation is 1 May 2026.

China’s decline is real, and larger than the coverage suggests

Chinese goods imports peaked at $562.9bn in the twelve months to September 2022. In the twelve months to May 2026 they were $264.1bn — a fall of 53.1%. Mexico, over the identical window, rose 26.5% to $558.1bn.

The two crossed in mid-2023. China led on a trailing-twelve-month basis for the last time in June 2023, and Mexico has led in every one of the 35 months since. In the single month of May 2026, Mexico shipped the United States $54,179.6m of goods against China’s $23,507.8m — more than twice as much.

US exports in the other direction fell further in proportional terms. They peaked at $157.7bn in the twelve months to April 2023 and stood at $105.5bn to May 2026, down 33.1%.

Mexico took 39% of what China lost

The nearshoring account of the last four years holds that production moved from China to Mexico. The direction is right. The magnitude is not.

Mexico absorbed 39.2% of China’s decline. Canada, routinely named alongside Mexico as a nearshoring destination, went the other way and fell $56.0bn. Japan was flat.

The residual is the finding. Countries outside those four added $393.0bn of annual goods exports to the United States — roughly one and a half times the entire remaining Chinese import bill. The substitution was real, it was large, and most of it did not go to the neighbours.

The tariffs are hard to find in import prices

The standard forecast was that tariffs would raise the landed cost of imported goods. The all-commodities import price index stood at 150.8 in June 2026, against 143.1 in September 2022 — up 5.4% in three years and eight months, below general inflation over the same period. Measured against January 2018, before the first tariff round, the index is up 19.2% across more than eight years.

That is a weaker effect than the tariff schedule alone would imply, and the index cannot tell you why. Three explanations are consistent with it — exporters absorbing duty in margin, sourcing shifting toward cheaper origins, and composition changing as dearer goods drop out of the basket — and this series cannot distinguish between them.

Chip output rose 71%. Chip prices fell 15%.

The specific claim that tariffs would raise semiconductor costs and squeeze electronics manufacturers is testable, and the data runs against it.

US semiconductor and electronic component production stood at 181.9 in June 2026 against 106.5 in January 2018 — up 70.8%, and up 35.7% since September 2022. Producer prices for semiconductors and related devices went the other way: 29.695 in June 2026 against 34.800 in January 2018, down 14.7%, and down 4.5% since September 2022.

Output up seventy per cent and prices down fifteen is the opposite of a cost-squeezed sector. It is also consistent with what the same series showed when chip output rose and prices fell last quarter, and with a record June for output that was not a seven-month climb.

What this data cannot settle

The $393.0bn is unattributed here. FRED carries no country series for Vietnam — IMPVN does not exist — and the same is true of most of the economies that plausibly absorbed the diverted volume. Naming them requires UN Comtrade’s bilateral detail, not this dataset. Anyone asserting a specific destination on the strength of the FRED country series is asserting more than the series contains.

Nor can these figures separate genuine relocation of production from transshipment, where goods of Chinese origin are routed through a third country before entering the United States. The customs data records the last country of shipment, not the origin of the value added inside the box.

The Stress Index reads 45.4 as of 28 July, inside the Normal band, and its global trade activity component sits at 10.4. A reordering of trade on this scale has not registered as trade stress — which is what you would expect if volumes were maintained while routing changed.

What to watch

The May-2026 gap between Mexico and China, currently $30,671.8m in a single month. It has widened in each of the last three annual comparisons. A narrowing would be the first evidence that the redirection has run its course.

The import price index (IR), currently 150.8, published with a 58-day lag. It is the cleanest available test of whether tariff costs reach American buyers. A move above the rate of general inflation would reverse the reading above.

US exports to China, $105.5bn and 33.1% below their April 2023 peak. The import side of this relationship is widely reported; the export side has fallen proportionally further and is watched far less.