Thematic: US-China decoupling — what has actually shifted versus the narrative, the three November expiries, and where the rent and risk really sit

2026-06-19 · long-form · v1

Short forms used in this file: [BRB] = Mauboussin & Callahan, The Base Rate Book (Credit Suisse, 2016); [CR] = Chancellor (ed.), Capital Returns (Palgrave, 2015); [WWYZ] = Wang, Wei, Yu & Zhu, "Measuring FDI and Trade-related Interdependence," NBER WP 34226 (2025).

Executive summary

The decoupling the tape prices is a bilateral trade story: China's share of US goods imports has fallen from roughly 22% in 2018 to 13.4% in 2024, and printed as low as 7.1% in May 2025 during the tariff spike — the lowest since 2001 T3. Read alone, that series says the two economies are pulling apart fast. The series is misleading on its own terms.

What has actually shifted is the routing, not the dependence. The best current measurement work — a 2025 NBER paper that combines trade data with firm-ownership and foreign-investment data — finds China's value embedded in US consumption rose from under 5% in 2000 to roughly 15-16% by 2020, about $20 billion to $210 billion, reaching near 44% in electronics and 56% in textiles T2. Direct China-US trade falls; the same Chinese value-added arrives through Vietnam, Mexico, and other connector economies, often inside the internal networks of the same multinational firms. Mexico alone accounts for about a quarter of China-linked inputs used in US production T2. Over $8 billion of Chinese exports were rerouted through Vietnam to the US in the first three quarters of 2025 T2. The headline decouples; the supply chain reconfigures.

The Friday read has three parts. First, the consensus mistakes a change in customs stamps for a change in economic dependence — the dependence is stickier than the trade data shows, which means the businesses most exposed to a genuine supply shock are not the ones the market has marked down. Second, the policy architecture is built almost entirely from suspended controls that expire together: the tariff truce, the semiconductor Affiliates Rule, and China's rare-earth control suspension all lapse around November 10, 2026. That convergence is the single most actionable calendar fact in the macro book and it is under-priced as a cluster. Third, the base rate for weaponized controls between deeply integrated economies is reversion, not permanence — Japan and South Korea restored their trade whitelists within four years; the 1980s US-Japan semiconductor managed-trade regime did not stop Japanese decline, which came from Korea and Taiwan instead. The durable winners are the connector economies and the genuine bottleneck holders, not the headline "China plays."

House view reconciliation

The house view carries critical minerals as a confirmed dossier (v1), framed as Phase 2 of a Marathon capital cycle with the duration of Chinese export controls as the load-bearing variable 2026-05-29-critical-minerals-capital-cycle-dossier-v1. It does not yet carry a standalone US-China decoupling theme. The AM and PM notes through June reference the pieces — the G7 Evian critical-minerals agenda, the November 2026 control expiry, semiconductor licensing — but as inputs to other files, not as a unified theme 2026-06-15-AM.

This dossier opens that theme and is deliberately scoped to not re-litigate critical minerals. Minerals is the materials leg and already has its own dossier; this is the umbrella over the trade, technology, and capital legs. The one place the two connect is the November 2026 expiry cluster, which this report elevates from a minerals-only calendar item to a cross-theme macro anchor. No conflict with the existing house view; this extends it. Specific edits at the end.

The setup

Decoupling became consensus in two waves. The first was the 2018-2019 Section 301 tariffs. The second was the 2025 escalation under the second Trump administration, when reciprocal tariffs briefly drove the effective US rate on China toward triple digits before the October 30, 2025 Busan meeting between Trump and Xi cut it back and extended a 10% truce rate to November 10, 2026 T3. A Supreme Court ruling on February 20, 2026 declared the IEEPA tariffs unconstitutional, after which the administration reimposed a 10% global tariff under Section 122 and moved to restore the China tariffs through Section 301 T3. As of mid-June 2026, China faces the highest effective tariff rate among major US trading partners at roughly 24%, with the overall US average effective rate near 7.0% T2.

That is the visible machinery. The question this report answers is narrower and more useful than "is decoupling happening": how much of the measured separation is real economic separation, and how much is rerouting that leaves the underlying dependence intact? The answer decides which businesses are actually de-risked and which only look de-risked because their Chinese content now arrives with a Vietnamese or Mexican label.

The analysis

What the trade data says, and why it overstates the case

China's share of US goods imports fell from about 22% in 2018 to 13.4% in 2024, and Mexico displaced China as the top source of US imports in 2023 for the first time in 17 years T3. The May 2025 print of 7.1% was a tariff-spike trough, not a trend level — front-loading and a brief near-embargo distorted that month — but even the smoothed series shows the steepest reallocation in the post-WTO era T3.

Two facts puncture the clean reading. First, China's own export data shows its share of the US import market falling by only about 2.5 percentage points, less than a third of the decline in the US figures — a gap consistent with transshipment and re-invoicing through third countries T3. Second, the connector economies that gained share are themselves importing more from China to make the goods they ship to America. In 2025, China's direct exports to the US fell about 20% while Vietnam's exports to the US rose 28% year-on-year T3. Trade did not leave the system. It changed address.

The reconfiguration mechanism — rerouting and the FDI channel

The measurement that matters is value-added, not gross bilateral flows, and the cleanest recent work is the Wang-Wei-Yu-Zhu NBER paper. It combines global production tables with firm-ownership data to separate trade-channel dependence from the dependence that runs through multinational investment networks T2. Three findings carry the argument.

China's value embedded in US consumption climbed from under 5% in 2000 to roughly 15-16% by 2020 — about $210 billion — and concentrates in exactly the categories where substitution is hardest: near 44% in electronics and 56% in textiles T2. A tariff that reroutes the iPhone's final assembly to India does little to the Chinese value-added inside it. Second, the indirect China inputs embedded in US production rose from under $5 billion in 2000 to over $30 billion by 2020, with Mexico alone carrying about a quarter of those China-linked inputs T2. Third, China's direct reliance on the US market fell only modestly, from nearly 20% to about 16-17%, even as its sales through Mexican, Canadian, Vietnamese, and Korean intermediaries expanded sharply T2. The paper's own summary line is the thesis of this section: this is "reconfiguration rather than true decoupling," and global production networks "remain resilient to policy shocks targeting direct bilateral flows" T2.

The transshipment work confirms the mechanism at transaction level. An average tariff hike of 12.5 percentage points on Chinese goods produced a 1.74-point increase in rerouting through Vietnam, a rise of over 14% versus pre-trade-war levels, with more than $8 billion rerouted through Vietnam in the first three quarters of 2025 T2. Washington has noticed: the 2025 framework added a 40% penalty on transshipped goods, concentrated on Vietnam, Malaysia, Thailand, Indonesia, and the UAE T3. That penalty is the tell that policymakers know the headline series is being gamed — and it raises the cost of the rerouting workaround without restoring domestic capacity.

The capital channel decoupled far more cleanly than the trade channel. Chinese FDI into North America fell to about 2.6% of China's total announced outbound investment in 2025, down from roughly 27% a decade earlier, even as China's global outbound FDI rose 18% to $124 billion T2. Capital and finance are separating; goods and value-added are merely re-routing. That divergence is the heart of the variant view.

The control regime is suspended, not resolved — three November expiries

The most important structural fact about the 2026 decoupling is that almost none of the de-escalation is permanent. It is a stack of one-year suspensions that lapse together.

The tariff truce from the October 30, 2025 Busan meeting holds the 10% rate to November 10, 2026 T3. The semiconductor "Affiliates Rule," which would extend export controls to foreign entities at least 50% owned by listed Chinese firms, was suspended on November 10, 2025 for exactly one year, to November 10, 2026 T3. China's October 2025 rare-earth processing and licensing controls were suspended for one year under the same November 2025 truce and expire in November 2026 [carried 2026-05-29-critical-minerals-capital-cycle-dossier-v1; T3: china-briefing.com]. Three separate control regimes — tariffs, US technology controls, Chinese mineral controls — all reset within days of each other in November 2026.

The direction of travel since then has been toward more trade, not less. On December 8, 2025 the administration cleared the H200 and AMD MI325X-class chips for approved Chinese customers, and on January 13, 2026 BIS moved their license review from "presumption of denial" to "case-by-case" T1. By GTC 2026 in March, Nvidia had purchase orders for H200 chips from Chinese customers and was restarting that production line T3. The controls are being used as bargaining chips, not as a wall — Washington added 28 Chinese firms to the Entity List in the same window it loosened the chip rule T1. Restriction at the firm level, relaxation at the product level. That is negotiation, not a permanent technological iron curtain.

The market is pricing each of these as separate, resolved-for-now headlines. The cluster is the risk. A breakdown in any one negotiation near November 2026 raises the conditional probability of breakdown in the others, because the same two governments hold all three and have shown they bundle them.

Where the rent and the risk actually sit

The decoupling narrative pushes capital toward visible "China-exposure-reduction" plays — domestic manufacturing, friend-shoring beneficiaries, the connector economies. The reconfiguration evidence says the exposure was not removed, so the screening question is which businesses hold a genuine bottleneck versus which are merely a new address for the same Chinese value-added.

The connector economies — Mexico, Vietnam — capture assembly margin and logistics rent, but the NBER value-added work shows they remain conduits for Chinese inputs, which means they carry the supply-shock risk without controlling the bottleneck and are now squarely in the path of the 40% transshipment penalty T2. The semiconductor equipment and design layer holds real leverage, because the controls have repeatedly targeted it and US-allied firms hold steps in the process that cannot be substituted on a short clock; the risk there is that relaxation toward China (the H200 reversal) compresses the scarcity premium the market had capitalized. The critical-minerals processing layer is the cleanest genuine bottleneck and already has its own dossier and discipline; nothing here changes the $50 central, $42 trigger read on MP Materials [carried 2026-05-29-critical-minerals-capital-cycle-dossier-v1]. The businesses the market treats as "de-risked from China" because their imports now say Vietnam are the ones most likely to be re-rated when a supply shock proves the dependence never left.

Base rates

Mauboussin's discipline is to favor the outside view — the distribution of outcomes for analogous situations — over the inside view built from the specifics of this case T2. The inside view here is "this time the great powers have decided to separate permanently." The outside view asks what happened the last several times integrated economies weaponized trade and technology controls against each other.

The closest recent analog is the Japan-South Korea dispute. On July 1, 2019, Japan imposed export licensing controls on three chemicals critical to Korean semiconductor and display production — hydrogen fluoride, photoresist, fluorinated polyimide — and removed Korea from its trusted-trade whitelist T3. Korea filed a WTO complaint and pushed domestic substitution. The dispute resolved after about four years: Japan restored Korea to the whitelist and Korea withdrew the WTO complaint in March 2023 T3. Base rate: weaponized controls between two deeply integrated, allied-adjacent economies reverted within four years, and the targeted economy diversified at the margin without achieving independence. US-China is less friendly and larger, which argues for a longer and messier path — but the direction of the base rate is reversion, and the 2026 H200 reversal and tariff truce are already consistent with it.

The deeper analog is the 1980s US-Japan technology conflict. After a 1985 Section 301 case and 1986 antidumping action, the US imposed 100% tariffs on selected Japanese electronics and signed managed-trade semiconductor agreements in 1986 and 1990 that set a 20% foreign market-share target in Japan T2. The friction ran roughly a decade. Two base-rate lessons carry. First, the managed-trade agreements did little to stop Japanese dominance directly — Japan's memory lead eroded instead because Korea and Taiwan built lower-cost capacity, a third-party outcome the bilateral framework never targeted. Second, the period that looked like permanent US-Japan economic decoupling at the time was followed by decades of deep integration. "This time is different" is the most expensive sentence in cyclical analysis T2.

Combining the analogs gives a usable base rate for the duration question. Weaponized-control episodes between integrated major economies have historically resolved or substantially de-escalated within roughly four to ten years; the targeted dependence reconfigures geographically rather than disappearing; and the eventual relative winner is frequently a third party that builds genuine low-cost capacity, not either principal in the dispute. Applied to 2026: the connector economies and bottleneck-capacity holders are the base-rate-favored winners; the permanence of the controls is the base-rate-disfavored bet; and the November 2026 cluster is a scheduled stress test of which way this episode breaks.

Variant perception

Consensus reads the falling bilateral trade share as evidence that US-China economic dependence is being durably reduced, and prices "China-exposed" businesses as structurally impaired while rewarding visible reshoring and connector-economy plays. The load-bearing assumption is that the trade data measures dependence. The NBER value-added work shows it does not: dependence fell far less than gross trade, the same Chinese value-added now arrives through Mexico and Vietnam, and the capital channel — not the goods channel — is what genuinely decoupled T2.

AlphaSteve's variant is two-sided. Near term, the decoupling is over-priced as a permanent regime — the policy is a stack of one-year suspensions trending toward relaxation, and the H200 reversal plus the 24% (not triple-digit) effective tariff are evidence the wall is negotiable T1. Multi-year, the reconfiguration is under-priced: the November 2026 expiry cluster is a real, scheduled, three-front risk that the market treats as three unrelated headlines, and the businesses marked "de-risked" because their imports relabeled to Vietnam still carry the original supply-shock exposure.

What would falsify the variant: a clean, durable extension of all three suspensions past November 2026 with tariffs trending lower would confirm the consensus de-escalation read and weaken the cluster-risk thesis. Evidence that domestic or allied capacity is genuinely displacing Chinese value-added — not just relabeling it — would mean the dependence is actually falling and the trade data is right after all. The NBER methodology applied to 2024-2026 data, if it showed China's embedded value-added in US consumption falling rather than holding, would break the core claim.

Implications for AlphaSteve

The top-down implication is to stop treating the China-import-share series as a dependence gauge and start treating the November 2026 expiry cluster as a single, scheduled macro event. The decoupling theme is real as a capital-and-policy story and over-stated as a goods-dependence story. The deep-value posture is to lean against the crowded "reshoring winner" trade where the de-risking is cosmetic, and to track the genuine bottlenecks and connector economies the base rate favors. No new position is initiated on this dossier; it sets the lens and the calendar.

• Portfolio: No position changes. This dossier is theme-level, not name-level.
• Watchlist: Add a "cosmetic de-risking" screen flag — names re-rated higher on reduced China import exposure whose value-added dependence (per supplier geography and bill-of-materials) is unchanged. These are short-side or avoid candidates if a supply shock re-prices the dependence.
• Theses on the workbench: Critical minerals (MP Materials) read is unchanged; the November expiry is now shared across the minerals and decoupling dossiers as a common catalyst, not a minerals-only one.
• Sectors: Semiconductors — flag that controls are loosening toward China (H200), which compresses the scarcity premium some equipment and design names carry; treat as a fade-the-control-premium watch, not a position. Connector economies (Mexico, Vietnam exposure) — conduit risk plus the 40% transshipment penalty; participant, not bottleneck.
• House view updates: Add a "US-China decoupling — reconfiguration, not separation" theme (dossier v1) to the Themes section. Note the three-front November 2026 expiry cluster as a standing macro calendar anchor cross-referenced from the critical-minerals dossier.
• Daily-scan adjustments: Add the November 2026 expiry cluster (tariff truce, semiconductor Affiliates Rule, China rare-earth control suspension) to the standing catalyst calendar. Add a value-added-versus-bilateral-trade check to any name screened on "China de-risking."

Charts / data

Table 1 — What decoupled and what only reconfigured (value-added view)

Table 2 — The three November 2026 expiries (the cluster)

Table 3 — Base rates for weaponized-control episodes between integrated economies

Sources

• T2 Wang, Z., Wei, S-J., Yu, X., Zhu, K. (2025). "Measuring FDI and Trade-related Interdependence Across Countries: The Case of the United States and China." NBER Working Paper 34226 — https://www.nber.org/papers/w34226. Summarized in [T2/T3] SCCEI China Brief, "U.S.-China Economic Interdependence Has Shifted, Not Disappeared," 2026-02-01 — https://sccei.fsi.stanford.edu/china-briefs/us-china-economic-interdependence-has-shifted-not-disappeared
• T2 Rhodium Group, China Cross-Border Monitor, 2025 — Chinese FDI into North America ~2.6% of total; global outbound +18% to $124B; via Rest of World, 2026 — https://rhg.com/china/data-and-tools/the-china-cross-border-monitor/
• T2 Penn Wharton Budget Model, "Effective Tariff Rates and Revenues (Updated June 16, 2026)" — China effective ~24%, US average ~7.0% — https://budgetmodel.wharton.upenn.edu/p/2026-06-16-effective-tariff-rates-and-revenues-updated-june-16-2026/
• T2 Harvard/Duke/Academia Sinica transshipment analysis, via The Diplomat, "Inside China's Rerouted Supply Chains," 2026-01 — https://thediplomat.com/2026/01/inside-chinas-rerouted-supply-chains/
• T2 Mauboussin, M. & Callahan, D., The Base Rate Book (Credit Suisse, 2016); outside-view formulation from Kahneman, D. & Lovallo, D., "Timid Choices and Bold Forecasts," Management Science (1993)
• T2 Chancellor, E. (ed.), Capital Returns: Investing Through the Capital Cycle (Palgrave Macmillan, 2015)
• T2 CSIS, "Japan and South Korea Turn the Page" — Japan-Korea dispute resolution — https://www.csis.org/analysis/japan-and-south-korea-turn-page
• T2 NBER, "The U.S.-Japan Semiconductor Trade Conflict," c8717 — https://www.nber.org/system/files/chapters/c8717/c8717.pdf; T2 Econofact, "Do Trade Restrictions Work? Lessons From Trade With Japan in the 1980s" — https://econofact.org/do-trade-restrictions-work-lessons-from-trade-with-japan-in-the-1980s
• T1 BIS, "Department of Commerce Revises License Review Policy for Semiconductors Exported to China," 2026-01-13 — https://www.bis.gov/press-release/department-commerce-revises-license-review-policy-semiconductors-exported-china; Entity List updates 2026
• T3 Bloomberg, "China Share of US Imports Is Back to Pre-WTO Levels," 2025-12-06 — https://www.bloomberg.com/news/newsletters/2025-12-06/china-share-of-us-imports-is-back-to-pre-wto-levels-new-economy
• T3 NY Fed Liberty Street Economics, "U.S. Imports from China Have Fallen by Less Than U.S. Data Indicate," 2025-02 — https://libertystreeteconomics.newyorkfed.org/2025/02/u-s-imports-from-china-have-fallen-by-less-than-u-s-data-indicate/
• T3 SCMP, "China dethroned as top source of US imports after 17 years, replaced by Mexico," 2024 — https://www.scmp.com/news/china/article/3251318/
• T3 Supply Chain 24/7, "U.S. Imports From China Fall to 13.4%, Lowest Share in Decades" — https://www.supplychain247.com/article/us-imports-from-china-record-low
• T3 China Briefing, "US-China Tariff Rates 2025/2026" — https://www.china-briefing.com/news/us-china-tariff-rates-2025/; T3 CRS IF12990, "U.S.-China Tariff Actions Since 2018"
• T3 Tax Foundation, "Tariff Tracker: 2026 Trump Tariffs & Trade War by the Numbers" — https://taxfoundation.org/research/all/federal/trump-tariffs-trade-war/
• T3 Mayer Brown, "Administration Policies on Advanced AI Chips Codified," 2026-01 — https://www.mayerbrown.com/en/insights/publications/2026/01/administration-policies-on-advanced-ai-chips-codified
• T3 Korea Herald, "Japan restores South Korea to export 'whitelist' after 4 years," 2023 — https://www.koreaherald.com/article/3156885
• T3 East Asia Forum, "Weaponising trade in the Japan-South Korea dispute," 2019
• See sources-policy for the citation discipline applied. Wikipedia excluded throughout.

House view changes this run

• Added to Themes section: "US-China decoupling — reconfiguration, not separation (dossier v1)." Rationale: a unified cross-sector theme distinct from the critical-minerals dossier, grounded in the NBER value-added evidence that bilateral trade overstates the separation, with an explicit near-term (over-priced permanence) / multi-year (under-priced reconfiguration and cluster risk) variant.
• Added standing macro calendar anchor: the three-front November 2026 expiry cluster (tariff truce, semiconductor Affiliates Rule, China rare-earth control suspension), cross-referenced from 2026-05-29-critical-minerals-capital-cycle-dossier-v1. Rationale: the market prices these as three unrelated headlines; the same two governments hold all three and bundle them.
• No name-level position changes. MP Materials read unchanged ($50 central, $42 trigger). No new positions initiated.
• Scan addition: value-added-versus-bilateral-trade check on any "China de-risking" candidate; November 2026 expiry cluster added to the catalyst calendar.

Linked

• _house-view
• 2026-05-29-critical-minerals-capital-cycle-dossier-v1 — the materials leg; shares the November 2026 expiry anchor
• 2026-06-05-ai-infrastructure-capacity-dossier-v1 — semiconductor controls intersect the AI-infra demand picture
• 2026-06-12-synchronized-tightening-energy-shock-v1 — prior Friday dossier
• capital-cycle · narrative-cycle · base-rates · sources-policy · voice-and-style