For the first time in decades, Taiwan has surpassed China as a source of US goods imports — not because of consumer electronics or furniture, but because of AI chips and servers flowing through TSMC’s fabs.
US Census data, corroborated by Taiwan News, Taipei Times, and Newsweek, shows that Taiwan shipped approximately $24.6 billion in goods to the United States in May 2026 — edging past China’s $23.5 billion for the same month. Over the January–May 2026 period, Taiwan’s cumulative total reached roughly $116 billion versus China’s $104 billion: the first sustained crossover in decades. A decade ago, Taiwan’s monthly figure sat near $3 billion. China’s peaked somewhere between $45 and $50 billion and has now collapsed to roughly half that.
Two forces are doing the work simultaneously. US purchases from China plunged approximately 44% year-on-year under the Trump-era tariff escalation — a policy-driven collapse in consumer goods, electronics assembly, and manufactured imports that China had dominated for a generation. At the same time, Taiwan’s advanced-system exports surged roughly 93.5%, driven by TSMC-fabricated AI silicon and the AI servers built around it. The composition of Taiwan’s export basket matters: this is not a broad-based manufacturing miracle. It is one category — compute — pulling the entire trade relationship upward.
A necessary hedge before the structural read: customs-value trade figures swing month to month and are sensitive to tariff-timing effects and front-loading behavior. “Imports from Taiwan” is a broad basket that includes many categories beyond chips. The AI hardware driver is real and documented, but it does not represent the entire $116 billion. The crossover is also recent and narrow — a single month’s spread of $1.1 billion is not a moat. What matters is the direction and the mechanism, not the precise dollar margin.
HOW THE TRADE MAP SHIFTED
~2015
Taiwan sends ~$3B/month to the US. China dominates at $40B+. The gap looks insurmountable.
2018–2022
Trump-era and Biden-era tariffs begin reshaping China’s export trajectory. AI capex buildout starts concentrating demand in advanced-node silicon — a TSMC specialty.
January 2026
US–Taiwan deal signed: Taiwanese firms commit $250B to US chip and AI investment. TSMC Arizona, Micron’s $250B memory buildout, and AI server supply chains accelerate.
May 2026
Taiwan ($24.6B) overtakes China ($23.5B) in monthly US goods imports for the first time in decades. Advanced-system exports up ~93.5% year-on-year.
The key insight: Taiwan did not diversify into new export categories — it dominated the one category that the entire US technology economy is now organized around. A single product class, advanced compute, is enough to reroute America’s largest import relationship. That is the measure of how large the AI supercycle has become in physical, dockside terms.
The Structural Read
The standard frame for this story is geopolitics: US-China decoupling pushes Chinese goods down, friend-shoring pulls Taiwan up. That frame is accurate but incomplete. The deeper mechanism is that trade now follows compute. Goldman Sachs has mapped approximately $1 trillion in cumulative AI infrastructure capex from the hyperscalers — and every dollar of that capex has to arrive somewhere as a physical object. The overwhelming share arrives as TSMC-fabricated silicon or as AI servers assembled around it. The trade ledger is the physical shadow of the capex curve.
This is not the first time a single technology has bent a trade relationship. Petroleum reshaped Middle East trade flows for a century. Semiconductors are doing something analogous in compressed time — except the geopolitical concentration is even more acute. Petroleum is distributed across dozens of producing nations. Leading-edge AI chips are fabricated in meaningful volume at essentially one location: TSMC’s advanced nodes in Taiwan. The AI map and the trade map are becoming the same map.
“When a $1 trillion capex buildout gets concentrated in a single geography’s manufacturing capacity, the trade data stops being economics and starts being a strategic dependency map. The crossover is not a milestone — it is a warning label.”
The same dynamic that explains the crossover also explains the rationing of Nvidia H200s inside China — Alibaba, ByteDance, and DeepSeek competing for a constrained supply of chips they cannot domestically fabricate. The decoupling that shows up on the US import ledger as “China out” shows up on the Chinese side as artificial scarcity of frontier compute. These are two faces of the same constraint. The geopolitical fencing of frontier AI is not a future risk scenario — it is the present operating environment, and trade statistics are now one of the most legible ways to read it.
The concentration risk cuts in both directions. The US AI infrastructure buildout — the one underpinning the Goldman $1T capex curve — is structurally dependent on an island sitting in a contested strait. The policy response is the reshoring race: TSMC Arizona, Micron’s $250B US memory buildout, and the January 2026 US-Taiwan deal under which Taiwanese firms committed $250 billion to US chip and AI investment. The reshoring race exists precisely because the trade crossover represents a dependency, not just a milestone.
Three Implications
IMPLICATION 1 — AI CAPEX HAS A PHYSICAL TRADE SIGNATURE
Every hyperscaler data center announcement, every GPU cluster, every AI server rack shows up eventually in customs data. The Taiwan crossover is the dockside face of the AI supercycle — and it means trade policy, export controls, and geopolitical disruption are now first-order risks to AI infrastructure timelines, not second-order considerations. Strategists at hyperscalers and AI labs need to read the trade ledger the way energy companies read oil shipment data.
IMPLICATION 2 — CONCENTRATION RISK IS NOW PRICED INTO POLICY, NOT YET INTO MARKETS
The January 2026 US-Taiwan $250B commitment and TSMC Arizona are explicit acknowledgments that a single-node dependency is unacceptable at strategic scale. But the reshoring timeline runs years to decades while the AI buildout is running quarters. The gap between when the dependency exists and when the redundancy is operational is the risk window — and it is open right now. The verticalization of AI stacks deepens this: the more hyperscalers own end-to-end silicon design through TSMC manufacturing, the more concentrated the single point of failure becomes.
IMPLICATION 3 — CHINA’S COMPUTE ISOLATION IS NOW VISIBLE IN TWO LEDGERS SIMULTANEOUSLY
US imports from China down 44% year-on-year. Chinese firms rationing H200s internally. These are not separate stories — they are the same supply-chain schism read from opposite ends. China is being squeezed out of both directions: it cannot export freely into the US market under current tariff regimes, and it cannot import the frontier chips it needs to run competitive AI infrastructure. The decoupling is now structurally symmetric, and it is accelerating. Any company whose AI strategy assumes a reunified global chip supply chain is building on a faulty premise.
Business Engineer Framework
Map of AI — The Trade Layer
The Map of AI framework tracks 200+ companies across 9 layers of the AI stack — from silicon fabrication through model training to application deployment. The Taiwan crossover is evidence that Layer 0 (physical compute infrastructure) is now the most geopolitically sensitive layer in the entire map, and understanding where that layer sits, who controls it, and what redundancy exists is the prerequisite for reading every other layer correctly. The AI Capex Map provides the financial architecture that sits on top of this physical layer.
Taiwan overtaking China in US imports is not a trade story with an AI subplot — it is an AI infrastructure story that happens to show up in trade data. TSMC’s advanced nodes are the physical bottleneck through which the entire AI
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
