The AI boom remains a key pillar of global trade resilience

Overview

This report examines the impact on global trade if the AI boom fades. HSBC points out that global merchandise trade has remained resilient over the past 18 months despite headwinds including fluctuations in US tariff policy and shipping disruptions in the Middle East, mainly because of strong demand for AI-related goods. AI-related goods currently account for nearly 20% of global merchandise trade, up significantly from an average of approximately 14% in 2024.

Core views

The core view is that global trade growth currently depends heavily on the AI cycle. AI-related goods and digital services will continue to support trade for some time, but non-technology goods exports have weakened noticeably since 2024. If the AI cycle cools, particularly if momentum in Asian hardware exports weakens, global export growth could slow accordingly. However, HSBC does not regard a sharp halving of export growth in Taiwan and South Korea as its base case.

Analysis framework

The report uses macro trade decomposition and scenario analysis, separating the performance of AI-related goods trade from non-AI goods trade. It combines export or import forecasts for major economies including Taiwan, South Korea, mainland China, and the United States to estimate the potential drag on global export growth from changes in the AI cycle. The report also references hyperscaler capital expenditure forecasts to assess the sustainability of AI demand.

Methodology notes

• Macro trade analysisDecomposition of AI-related goods' contribution to trade

  AI-enabling goods are identified separately from global merchandise trade, and their share, contribution to growth, and regional distribution are compared.

  This framework is used to determine whether global trade growth is primarily driven by AI-related goods and to identify the exposure of economies including Taiwan, South Korea, mainland China, Hong Kong, and the United States to the AI value chain.

• Scenario analysisScenario of halved export growth

  Assume that Taiwan and South Korea's export growth forecasts are halved in 2026 and 2027, and measure the impact on global export growth.

  This scenario is not HSBC's base-case view; it is used to measure the sensitivity of global trade growth to a cooling AI hardware export cycle.

• Capital expenditure trackingHyperscaler capital expenditure forecasts

  Monitor capital expenditure trends at AWS, Google, Oracle, CoreWeave, Microsoft, and META.

  The report uses these companies' forecasts for data center and AI infrastructure investment to assess whether demand for AI-related components still has sustained support.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Asian AI hardware export chain

  Directly benefits from the expansion of AI-related goods trade

  Strengths

  Taiwan, South Korea, mainland China, and ASEAN economies occupy important positions in hardware, components, and key inputs.

  Weaknesses

  High dependence on the AI cycle, weak non-technology export growth, and potentially amplified volatility from concentration of demand.

  Comparison

  Compared with non-AI goods, AI-related goods have grown significantly faster since 2024.

  Risks

  Reversal of AI investment, decline in component prices, and an increase in non-tariff measures.

• US imports and data center investment

  The United States is both an important destination for AI-related imports and a core region for data center capital expenditure.

  Strengths

  Approximately 40% of data centers are located in the United States, and hyperscaler capital expenditure is expected to remain strong.

  Weaknesses

  The share of AI-related goods in US imports is rising rapidly, with some of the growth driven by prices rather than volumes.

  Comparison

  The growth in the value of US AI-related imports is significantly higher than the growth in import volumes.

  Risks

  Slower capital expenditure growth, changes in tariffs and non-tariff measures, and a reversal in the technology investment cycle.

• Digitally delivered services

  AI could boost services trade through productivity gains and lower trade costs.

  Strengths

  Digitally delivered services already account for more than half of global services exports, and the WTO model forecasts AI-driven growth of 39%-42% from 2025 to 2040.

  Weaknesses

  Realization of growth depends on the diffusion of AI applications, cross-border services regulation, and infrastructure investment.

  Comparison

  The report describes this as potentially one of the services sectors most boosted by AI.

  Risks

  Regulatory restrictions, cross-border data barriers, and a slowdown in AI investment.

Key data

• Share of AI-related goods in global merchandise tradeNearly 20%Up from an average of approximately 14% in 2024.
• Contribution of AI-related goods to last year's global merchandise trade growthMore than 40%The report cites WTO figures.
• Contribution of AI-related goods to nominal export growth in the first quarter of 2026Approximately 80%HSBC's estimate; note that the nominal measure was affected by rising component prices.
• Share of Taiwan's exports related to AIApproximately 80%Shows Taiwan's high dependence on AI value-chain exports.
• Share of US imports related to AIApproximately 27%The share has nearly doubled since early 2025, mainly driven by higher component prices.
• Scale of digitally delivered services exportsApproximately USD 5.3 trillion in 2025, accounting for 55% of global services exportsThe report states that digitally delivered services exports grew 10% year on year in 2025.
• Taiwan export forecastApproximately +23% in 2026 and +12% in 2027Forecasts by HSBC economists covering South Korea and Taiwan.
• South Korea export forecastApproximately +7.8% in 2026 and +3.1% in 2027Forecast for trade in goods and services on a real basis.
• Impact under the scenario analysisA drag of approximately 0.2-0.3 percentage points per year on global export growthAssumes that Taiwan and South Korea's export growth forecasts are halved in 2026 and 2027.
• Capital expenditure of the six largest hyperscalersCould exceed USD 1 trillion in 2027Even if year-on-year growth slows to approximately 25% in 2027, the absolute scale would remain strong.

Impact & implications

The investment implication is that the short-term resilience of global trade, Asian exports, and some digital services remains closely tied to the AI capital expenditure cycle. If demand for AI hardware or data center investment slows, the related high-tech export chains in Taiwan, South Korea, and mainland China could be affected first, further weighing on global export growth. Conversely, as long as hyperscaler capital expenditure remains elevated, trade in AI-related components and digital services may continue to receive support.

Risks

• The AI investment boom could ultimately reverse, as did historical canal booms and the internet bubble, triggering a broader economic slowdown.
• Non-technology exports have been broadly stagnant since 2024, indicating that the broad foundation of global trade is not stable.
• Economies such as Taiwan and South Korea have high dependence on AI hardware exports; if the AI cycle cools, export growth could come under significant pressure.
• Tariffs on AI-related goods remain low, but non-tariff measures are increasing and could weaken trade flows.
• Some of the growth in AI-related trade reflects higher component prices; if prices decline, nominal trade growth could slow.

What to watch

• Whether export orders and the growth of semiconductor and AI hardware exports in Taiwan and South Korea slow.
• Whether mainland China's high-tech export growth can remain strong.
• Whether the gap between the value and volume of US AI-related imports narrows.
• Whether AWS, Google, Oracle, CoreWeave, Microsoft, and META lower their capital expenditure guidance.
• Whether tariffs and non-tariff measures on AI-related goods increase further.
• Whether growth in digitally delivered services exports continues to be supported by the diffusion of AI applications.