Home / Guides / Sourcing Shift Evidence

US Imports From China Fell 22% While Vietnam Rose 39%: The Sourcing Shift, With Receipts

Computed from published UN Comtrade records (USA reporter, imports, TOTAL level, H1 2025 vs H1 2026); every figure below is reproducible from the sources listed in the measurement trail at the end of this guide.

Two stories about tariff-driven sourcing dominate trade commentary. One says China has been priced out of American supply chains. The other says nothing has really moved and the headlines are noise. Both cannot be true, and most commentary offers neither table nor method — just adjectives. We pulled the actual monthly records for both countries across two years and put them side by side. The data supports a third story: a genuine, large reallocation whose pace is now visibly changing.

The headline halves: one origin down 22.3%, another up 39.1%

Source: UN Comtrade public preview API, USA reporter (842), imports, HS TOTAL; probe markers [A1]/[A2]/[A4], consistency gates [G1]/[G2] 12/12 months each.

A 61.4-point momentum spread between the two origins [A3] is not a rounding artifact or a seasonal wiggle — it is the largest half-on-half rebalancing we could reconstruct at this level of granularity, and it happened while the combined basket barely moved. Buyers did not stop buying. They changed where they buy.

Six month pairs, no interpolation

Aggregates hide turning points, so here is every month-to-month comparison. Each cell is the same calendar month, one year apart — no seasonality adjustment, no estimated values:

Source: month-level values from the archived bilateral series, cross-checked file-vs-file 12/12 months per country [G2]; computations [A6].

Read the sign pattern, because it carries the news: China's year-over-year reading is negative in four of six months, Vietnam's positive in all six [A7]. But the two most recent months break the simple narrative — China's May and June 2026 imports came in above their year-earlier levels. Whatever your position on tariffs and supply chains, "China flat-lines to zero" is not what the record shows. The realistic reading is front-loaded disruption early in 2026 followed by partial re-stabilization, which matches how importers actually behave once new-measure costs settle into contracts.

Composition shifted 14 points while the total barely moved

Vietnam's share of the combined China-plus-Vietnam basket rose from 34.1% in H1 2025 to 48.0% in H1 2026 — a 14.0-point gain in the year-over-year comparison [A5]. Within 2026 itself, both origins grew quarter-on-quarter into Q2 (China $64.97B → $72.98B, +12.3%; Vietnam $57.88B → $69.66B, +20.4%), so the second quarter set up a tighter race than the first [A8]. If you plan production or inventory against a "China shrinking" mental model, you are planning against last winter's snapshot, not the current one.

Method, and what this data cannot tell you

Honesty about limits is what makes the numbers above usable:

• Level of aggregation. These are TOTAL-level flows. They answer "how much, from whom" and stay silent on which products moved. Product-level follow-ups need chapter or six-digit queries, which the same public API serves one call at a time — for a sample of the per-line view, see HS chapter 84 (machinery).
• Values, not units. Dollar values can move with prices as well as quantities. Nothing here distinguishes "fewer containers" from "cheaper containers".
• Mirror reporting. Partner-side records can differ from US-side records for the same physical flow; we use the US-reported series consistently for every figure.
• Capture window. Records were collected 2026-10-13T23:46Z through 2026-10-14T00:56Z with rate-limit retries, then re-verified by an independent fresh pull that matched all 24 month-values exactly [P3][G2]. Later revisions by statistical offices may adjust individual months.

For readers who want to go deeper themselves: the query parameters, the archived month-values and the verification verdicts are all linked from the measurement trail below. If your work needs the product dimension — which specific chapters carry the rebalancing — start with our background pieces on Section 301 China tariffs, the fabric-to-apparel tariff ladder plus the per-line rate tables for HS chapter 61 (knitted apparel) and HS chapter 62 (woven apparel), and what changed for small parcels after the $800 de minimis suspension. The macro picture also appears in our October 2026 tariff landscape report.

From macro shift to your invoice

A 61.4-point spread between two origins is a market-level statistic; your own decision is always narrower and sharper. The same product declared from different origins can face entirely different Chapter 99 exposures, and origin rules are exactly where margin quietly leaks. Model the alternatives in the duty calculator before committing a purchase order, browse origin-side infrastructure such as Vietnamese port profiles or Chinese port profiles, and if you need the moving picture rather than a snapshot, set up measure-change alerts or compare what monitoring includes on the pricing page.

Frequently asked questions

Is the United States really importing less from China?

In the first half of 2026, US imports from China totaled $137.94 billion against $177.42 billion in the same half of 2025 — a decline of 22.3% computed from UN Comtrade records (USA reporter, imports, TOTAL level). The decline is concentrated in January–April 2026; May and June 2026 actually rose above their year-earlier months.

How fast have US imports from Vietnam grown?

First-half US imports from Vietnam reached $127.54 billion in 2026 versus $91.68 billion in 2025, a 39.1% increase. All six month-over-month comparisons were positive, ranging from +31.4% (February) to +50.2% (January).

Is this a sourcing shift or just cheaper trade?

The combined China-plus-Vietnam basket moved only 1.3% lower ($269.11B to $265.48B), while Vietnam's share of that basket jumped from 34.1% to 48.0%. A roughly flat total with sharply changed composition is the signature of reallocation between origins rather than a collapse in demand.

Which months show the sharpest changes?

January shows the widest gap: China fell 48.7% year over year while Vietnam rose 50.2%. The gap narrows through spring, and by May–June 2026 China's year-over-year readings turned positive (+15.1% and +32.8%) while Vietnam posted +35.8% and +41.6%.

Where can I reproduce these numbers?

Query the UN Comtrade public preview API with reporterCode 842 (United States), flowCode M (imports), cmdCode TOTAL, periods 202501–202606, partnerCodes 156 (China) and 552 (Vietnam). Our archive stores the exact monthly values used here plus an independent verification pass that re-pulled both partners and matched every figure.

What does the shift mean for tariff planning?

Origin decisions now move landed cost as much as product category does: Chapter 99 measures such as Section 301 actions apply by origin, so the same product can carry very different totals depending on where it ships from. Model origin scenarios in the duty calculator before committing purchase orders, and monitor measure updates by subscription.

Measurement trail (test-before-write) — quantitative claims computed 2026-10-15 by .state/scripts/probe_sourcing_shift_1015.py; raw output archived at .state/media/output/article-sourcing-1015/probe-output.txt. Marker map: [A1] China H1'25 $177.42B → H1'26 $137.94B = −22.3%; [A2] Vietnam $91.68B → $127.54B = +39.1%; [A3] spread 61.4pt; [A4] combined $269.11B → $265.48B = −1.3%; [A5] Vietnam share 34.1% → 48.0% (+14.0pt); [A6] six month-pair rows as tabulated; [A7] sign patterns 4/6 negative (China) and 6/6 positive (Vietnam); [A8] Q1→Q2 2026 QoQ China +12.3% / Vietnam +20.4%; [G1] stored half-year sums equal sum-of-months (both countries, both years); [G2] archived months equal independent fresh-pull raw 12/12 per country; [P3] collection window 2026-10-13T23:46Z–2026-10-14T00:56Z with verify verdict China −22.3% PASS / Vietnam +39.1% PASS / spread 61.4pt PASS / 12+12 months. Reproduce: python3 .state/scripts/probe_sourcing_shift_1015.py (output archived alongside this draft). Primary source: UN Comtrade public preview API (comtradeapi.un.org, reporter 842, flow M, classification HS, cmdCode TOTAL, partners 156/552), redistribution per its Terms of Use with attribution; archive files topic-pool-revive-1013/comtrade_bilateral_h1.json and verify_fresh_pull_raw.json. Largest-half-rebalancing phrasing is scoped to this dataset (two partners, TOTAL level, two half-years) and is not a market-wide superlative.