Cross Asset Lens

The Headline Export Boom Is Real, but the AI Explanation Is Not

Fact
China’s nominal exports rose 18.5% year to date through July 2026, the fastest pace since the
2021 reopening surge. Export volumes were also strong, increasing 10.8% in the first half. The
question is not whether exports accelerated. It is what actually produced that acceleration and how much of it
represents value created inside China.

A narrow group of AI-related products – semiconductors, printed circuits, storage and processing units, and
computer parts – supplied almost half of nominal export growth in the second quarter. Downstream electronic
devices are excluded from this definition, so the measure is intended to isolate the hardware used to expand AI
capacity rather than capture digital products generally.

Interpretation
The apparent AI boom becomes much less powerful once price and volume are separated. AI-related export volume
growth fell to -1.4% year over year in Q2, while these products generated roughly two-thirds of
the recent rise in export prices. Memory prices had increased five- to sevenfold from mid-2025 as data-center
demand collided with limited supply. That is a global hardware-pricing shock flowing through China’s customs
data, not evidence that AI units produced in China suddenly became the principal engine of real export growth.

AI exports overstate the value retained inside China

Fact
Around 80% of nominal AI-related exports are shipped under processing-trade arrangements or through special
customs supervision zones rather than ordinary trade. These systems typically import components, assemble or
integrate them in China, and then re-export the finished product. Semiconductor supply chains are especially
fragmented, and foreign memory producers operate major plants inside China.

The customs value records the full product crossing the border, not the portion created domestically. The
AI-related trade balance therefore remains weak despite the large gross export number. This suggests that the
near-term beneficiaries are narrower than the headline implies: assembly, selected manufacturing capacity and
supply-chain activity may gain, but the export value should not be read as equivalent domestic profit or income.

The real engine is electrical machinery, transport and the New Three

Once price effects are removed, ex-AI electrical equipment and machinery remain the largest contributors to real
export growth. Solar cells, electric vehicles and lithium-ion batteries – the established “New Three” –
account for only about 5% of nominal exports, yet recently added roughly 1.5-2 percentage points to real export
growth. AI-related goods contributed less than one percentage point, and that contribution faded as volume growth
softened.

Interpretation
The more durable investment thesis is China’s ability to sell increasingly sophisticated manufactured goods at
competitive cost across several industries. The export cycle is not resting on one semiconductor category.
Electrical machinery, vehicles and selected chemicals provide a wider base, while the New Three deliver a much
larger volume contribution than their small export share would suggest. This favors businesses with localized
supply chains and demonstrable unit growth over those benefiting mainly from elevated component prices.

The European market is already part of that volume story. Real exports of electrical machinery and transport
equipment to the EU increased 19% in the first half of 2026. That strength explains why Europe has become the
most important marginal policy risk: the sectors gaining share are also the sectors creating the greatest
political pressure.

Emerging-market demand is more genuine than the rerouting narrative suggests

Chinese exports to emerging markets are often interpreted as goods ultimately destined for developed economies.
The value-added data tell a more nuanced story. Emerging markets received around 39% of China’s gross goods
exports and absorbed roughly 38% of the Chinese value added embedded in foreign final demand. Consumption
represented about 60% of that final demand, indicating that local households – not only industrial relocation or
transshipment – are a major source of demand.

ASEAN remains the clear exception and the main active rerouting hub. It received about 12% of China’s gross
exports but absorbed only 9% of Chinese value added in final demand. Even there, however, roughly three-quarters
of shipments still reflected local demand, particularly for electric vehicles and capital goods. Mexico showed a
smaller gap and did not display the same transshipment pattern during the 2025 trade conflict that it had during
the 2018 episode.

Hidden implication
China’s export diversification has become more resilient than a simple tariff-evasion story would imply. Genuine
consumption and investment demand in emerging markets reduce dependence on any single developed market. ASEAN
should still be monitored separately because tighter enforcement against rerouting could affect the portion of
trade that is not locally absorbed.

The New New Three are strategic options, not current export engines

AI, robotics and innovative drugs have been designated as the next generation of strategic industries. Their
potential is real, but their measurable contribution remains tiny and their economics differ sharply from the
goods-led New Three.

• AI services: Chinese providers represented 71% of global token consumption on one developer-oriented platform in July, but the platform is skewed toward individual users and many models are free or priced very cheaply. Usage can therefore grow much faster than cross-border revenue.
• Industrial robots: China became a net exporter in 2025, yet total robotics exports were only US$747 million in the first half of 2026, equal to about 0.03% of total exports.
• Innovative drugs: Announced outbound licensing deal value rose 66% year over year to roughly US$100 billion in the first half, but payments depend on clinical, regulatory and commercial milestones and will arrive gradually, if at all.

Conditional
These sectors could eventually shift China’s export mix from physical goods toward intellectual property,
software capability and licensing income. For now, investors should track realized service revenue, milestone
receipts and repeat commercial demand rather than headline model usage or announced transaction value.

Imports confirm that external strength is not a broad domestic recovery

Nominal imports increased 26.6% in the first half of 2026, but import volumes rose only 7.8%, with real growth
concentrated in gold. Energy volumes fell as high prices and ample inventories reduced purchases, while AI-related
imports largely reflected processing-trade and supply-chain needs. Gold flows were also shaped by quotas under a
tightly managed import regime, making them an imperfect measure of private demand.

Excluding gold, the underlying trade surplus widened to approximately US$1.3 trillion over the prior 12
months by July. The forecast embedded in the analysis calls for export-volume growth of 7.9% in 2026,
including 5.3% in the second half, followed by about 5% in later years. Import-volume growth is projected at 4.4%
in 2026 and roughly 2.5% thereafter.

Interpretation
The macro message is an expanding gap between external production strength and domestic absorption. Investors
seeking exposure to China’s export machine should not assume that the same data confirm a broad rebound in
domestic demand. Export manufacturers and domestic-demand assets are being driven by different mechanisms.

Europe Is the Main Constraint, but 27% Coverage Is Not 27% Lost

At the report date, the effective US tariff rate on Chinese imports had stabilized near 23%, shifting attention
toward Europe. The EU represented 15% of China’s exports in 2025, and its policy response is becoming more
assertive as Chinese electrical machinery, vehicles and computers gain share.

Existing and proposed European measures could touch, at least at the margin, around 27% of China’s annual exports
to the EU. The most important proposals include countervailing duties of roughly 17%-38% on plug-in hybrid
vehicles and an extension of the Carbon Border Adjustment Mechanism to downstream products. That extension could
add US$58 billion of exports to the mechanism and raise the affected share of China’s EU exports from 2.1% to
12.1%.

Interpretation
The 27% figure measures potential policy reach, not an expected export loss. Downstream carbon charges would apply
to the embedded emissions of steel and aluminum inputs rather than the entire value of finished machinery, and
implementation is not expected before 2028 at the earliest. The real impact will depend on product design, carbon
intensity, final duty rates and exporters’ ability to shift their product mix.

Chinese vehicle exporters have already shown that adaptability. Battery-electric vehicle shipments recovered in
2026 after an initial decline under existing duties, while plug-in hybrids and internal-combustion vehicles rose
rapidly because they were outside the additional battery-electric measures. Cost advantages, improving product
quality and China’s supply of more than 90% of the EU’s rare-earth elements by weight also limit Europe’s freedom
to escalate without commercial consequences.

Beyond AI Hardware: Machinery, EVs, Batteries and ASEAN Supply Chains

AI Export Volumes, the “New Three” and EU Trade Measures

EU Trade Barriers and Slower Emerging-Market Demand Could Stall Exports

• European policy becomes broad rather than targeted. The assumption of manageable friction weakens if anti-dumping duties, PHEV tariffs and downstream carbon charges expand quickly across the full machinery and transport complex.
• Emerging-market consumption slows. The diversification thesis depends on genuine local demand; a synchronized EM slowdown would expose how much growth still relies on external absorption.
• ASEAN rerouting faces stricter enforcement. Even though most ASEAN-bound trade is locally absorbed, tighter controls could disrupt the remaining transshipment channel and associated supply chains.
• Manufacturing cost advantages erode. Higher costs or stalled product-quality gains would challenge the central claim that broad competitiveness can offset trade barriers.
• Second-half volumes miss the forecast path. Failure to achieve the implied 5.3% H2 export-volume growth would indicate that the apparent breadth of the cycle was less durable than expected.

The evidence also has measurement limits. The AI definition is deliberately narrow, processing trade complicates
ownership of value added, input-output tables are backward-looking, and services such as model usage and drug
licensing are not captured cleanly. The article’s strongest conclusions concern the direction of the mechanisms,
not precise company-level earnings sensitivity.