• AI goods trade grew 29% in 2023–25, far outpacing overall trade
• AI hardware trade concentrated in a handful of economies
• Asia’s output would fall 1.7% below forecast in a correction

A sharp correction in artificial intelligence (AI) investment would cut Asia’s economic output by 1.7% below forecast, a hit second only to that of the United States among the regions the International Monetary Fund (IMF) reports, the fund warned on Tuesday (13). The IMF also noted that the AI boom, which has powered world trade, rests on supply chains concentrated in a handful of economies.

In a risk scenario in its October 2026 World Economic Outlook (WEO), released at the IMF-World Bank Group Annual Meetings in Bangkok, the IMF assumes the expected productivity gains from AI fail to materialise. In that scenario, AI investment unwinds and US equity prices fall by 20% in 2026, with prices elsewhere falling by 15%. Global economic activity would then be 1.1% below the baseline forecast by 2028, with larger losses in the United States, at 1.8%, and in Asia, at 1.7%.

Asia’s exposure stems from its role as the main source of AI-related capital goods. The report says Asian technology exporters would see export prices fall by 20% in such a correction, while other economies in the region would see a 4% decline. Emerging markets would also face temporarily higher borrowing costs, with corporate risk premiums and sovereign term premiums both rising by 75 to 100 basis points.

The scale of AI’s role in trade was set out at the WEO press briefing by IMF Economic Counsellor and Director of Research Silvana Tenreyro. She said global trade grew by 8.3% cumulatively between 2023 and 2025, while trade in AI-related goods grew by 29% over the same period. Without AI-related products, she said, overall trade growth would have been only 4.6%.

“These are big numbers,” Tenreyro said.

The gains have been unevenly spread. According to the report, Taiwan and Korea account for 62% of net trade in AI-related computing hardware. China and the United States accounted for more than 75% of net imports of such hardware in 2025, and have hosted 57% of new data centres built since 2024.

Tenreyro said AI supply chains were “highly concentrated in a few countries, sectors and firms, which creates potential choke points.” She added that a market value correction and AI-related threats to cybersecurity could weigh on growth and add to financial stability risks.

The report also questions whether the boom is yet paying off in productivity. Experiments show generative AI raising output in specific tasks such as customer support and software development. But the IMF said large surveys and official data show widespread adoption with limited effects on wages, hours worked, employment or measured productivity so far.

The IMF said the boom is also contributing to inflationary pressures and financial fragility.

Spending on semiconductors, data centres, power generation, and cooling systems is raising demand for capital goods, energy and skilled labour before broader productivity gains have materialised. Spillovers to core inflation in major economies appear limited so far. Rising technology stocks have supported spending through wealth effects, the report said, but at the cost of more stretched equity valuations and greater market concentration among AI-related

firms. It added that the calm in financial markets despite repeated shocks may partly reflect AI-related exuberance.

The fund’s baseline forecast assumes the AI boom gradually fades. In a more favourable scenario, where the boom lasts longer and delivers real productivity gains, world output would be 0.6% above the baseline by 2029. Output in Asia excluding China would peak 0.8% above the baseline, with most of the gains going to high-end technology exporters. In the long-term, the IMF said, the smallest gains would go to emerging markets least exposed to automation and technology industries.