In the medium to long term, AI can promote trade by reducing trade costs and enhancing productivity, while in the short term, it drives exports through AI-related products.
The WTO estimates that from 2025 to 2040, global trade volumes are projected to increase by 33.7% to 36.7% under various scenarios, with manufacturing trade volumes expected to grow by 22.2% to 24.4%. Key supporting factors include three main categories: first, the reduction of trade costs; second, the improvement of AI services and tradability; and third, the enhancement of productivity in tradable sectors. In the short term, AI primarily boosts exports through AI-related products, which can be classified into three major categories: first, raw materials and processed chemicals such as silicon dioxide, germanium oxide, zirconium dioxide, and silicon carbide; second, intermediate goods used in semiconductor production; and third, AI-related equipment, including computers, semiconductors, and related machinery [1].
AI-related product exports are accelerating.Based on the WTO’s list of AI-related products and aggregated data from economies with available statistics in UN Comtrade, global exports of AI-related products reached $3.1 trillion in 2024, a 10.1% increase from 2023. This accounted for 14.5% of total global exports, up 1.2 percentage points from 2023. In terms of export distribution by region, mainland China ($633.6 billion), Hong Kong, China ($355.4 billion), and Taiwan, China ($310.6 billion) ranked among the top three. The share of AI-related product exports in mainland China’s total exports was 17.7%, relatively low among major economies. In mainland China’s AI-related export structure, intermediate goods remain predominant, but the distribution between intermediate goods and equipment is relatively balanced compared to other major economies, accounting for 68% and 31%, respectively. For imports by region, mainland China ($541.3 billion), the United States ($479.5 billion), and Hong Kong, China ($346 billion) were the top three. However, in terms of AI-related equipment imports, the United States ranked first in 2024, with $201.6 billion.
U.S. AI capital expenditure drives imports of AI-related products.Amid the surge in AI investment in the United States, private-sector investment in information processing equipment has increased significantly, rising 26% year-on-year in the second quarter of 2025. This has also driven growth in imports of AI-related products, with U.S. imports of such products up 27% year-on-year in August 2025. Although this marks a decline from earlier peaks, it remains at a relatively high growth rate. The share of AI-related product imports in total U.S. imports has also continued to rise, reaching 20.7% in August 2025, an increase of 5.5 percentage points from the same period the previous year. Notably, the proportion of AI-related equipment imports has risen more significantly.
Mainland China’s AI-related product exports are steadily increasing, but the marginal boost remains relatively moderate.
From December 2024 to November 2025, mainland China’s exports of AI-related products amounted to $690.3 billion, a 9.7% year-on-year increase, though the absolute value remained below the level of July 2022. In terms of the contribution of AI-related products to mainland China’s monthly export growth year-on-year, while there has been a marginal increase, the extent remains relatively modest. From January to November 2025, AI-related products contributed approximately 1.7 percentage points to China’s export growth year-on-year, an increase of 0.3 percentage points from 1.4 percentage points in 2024. The contribution rate was 32%, up from 24% in 2024, but still lower than the WTO estimate [2] of AI-related products contributing 43% to global trade growth in the first half of 2025. Compared to other major economies, from January to October 2025, export growth in Taiwan, China, Malaysia, Singapore, and Thailand was almost entirely driven by AI-related products, while exports from regions such as South Africa, Brazil, Canada, and Australia were largely unaffected by AI-related products. On the other hand, direct trade in AI-related products between China and the United States is not substantial. From January to September 2025, the main sources of U.S. AI-related product imports were: Taiwan, China (24%), Mexico (22%), and Vietnam (12%), with mainland China accounting for only 8%. From January to October 2025, the main destinations for mainland China’s AI-related product exports were: Hong Kong, China (22%), the United States (10%), and Vietnam (8%). Therefore, in this sense, even if U.S. AI capital expenditure growth slows in 2026, its direct impact on mainland China’s overall export growth is likely to be relatively limited.
Exports of AI-related services are also accelerating.AI-related services are primarily included in telecommunications, computer, and information services within trade in services. Based on the rolling 12-month data for China’s exports and imports of telecommunications, computer, and information services, China’s exports of AI-related services are also likely showing an accelerating upward trend. As of October 2025, the rolling 12-month export value reached $70.8 billion, a 13.5% year-on-year increase, higher than the 9.6% growth for AI-related products. In a cross-country comparison, the regions with the highest export values for telecommunications, computer, and information services in 2024 were Ireland ($301.7 billion), India ($177.7 billion), the United States ($90.8 billion), China ($65 billion), and the United Kingdom ($59.9 billion).
China’s exports of AI-related services may be more oriented toward developing economies.
Due to differences in comparative advantages in AI development between China and the United States, China’s AI-related services may gain a competitive edge in developing economies by offering lower unit usage costs. According to a WTO study [3], open-source AI models used in low-income economies are predominantly from China, accounting for 81%, while those from the United States account for only 9%. In high-income economies, open-source AI models from China account for 34%, while those from the United States account for 59%.
