Vietnam's economy grew 7.4% in 2026, the largest upward revision in the region. Malaysia grew 5.1%. Thailand grew 2.0%. Pacific Island economies grew 2.2%, half a percentage point below their previous forecast, weighed down by energy costs they cannot absorb. The World Bank's October 2026 East Asia and Pacific Economic Update carries all of these numbers in the same document. What it does not do is describe them as part of the same story. They are not. The countries growing fastest are the ones plugged into the global AI hardware supply chain. The ones growing slowest are the ones that are not, and whose energy costs are rising precisely because the data centres, semiconductor plants and electronics factories elsewhere in the region are consuming more power.

The headline number in the World Bank report is $1.4 trillion. That is the value of AI-related goods exported by six economies -- China, Indonesia, Malaysia, the Philippines, Thailand and Vietnam -- in the twelve months to April 2026. The United States alone imported $739 billion of AI-related goods in that period, one third of it from East Asia and the Pacific. Gross AI-related capital expenditure reached 6% of American GDP. Of the $2.9 trillion in AI infrastructure investment planned globally for 2025 to 2028, $800 billion is expected to be financed through private credit. AI-related lending accounted for 34% of all private credit activity in 2025, up from an 18% average in the prior five years, according to the Financial Stability Board.

The specifics of which countries are benefiting reveal the precise nature of the wave. In Malaysia, AI-related goods contributed 51.5 percentage points of export growth against 19.7 percentage points from everything else. In Vietnam, 31.5 points against 6.7. In Thailand, 26.5 against 7.9. In the Philippines, 17.2 against negative 0.1 -- meaning that without AI-related goods, Philippine export growth was essentially flat. The composition of those exports matters. Assembled equipment -- computers, servers, routers -- made up 69% of Thailand's AI-related exports and 60% of Vietnam's. These are not AI companies. They are precision assembly economies plugged into a global hardware value chain whose final customers are in the United States and Europe.

Indonesia is the outlier in this picture and deserves attention. The region's largest economy by population -- 280 million people -- contributed just 0.5 percentage points of AI-related export growth against 4.6 points from conventional goods. Indonesia is not plugged into the semiconductor and electronics assembly chain in the way Malaysia, Vietnam and Thailand are. Its export growth comes from commodities and manufactured goods. Whether Indonesia's absence from the AI supply chain is a structural problem or a missed positioning decision is one of the more consequential political economy questions in the region.

Pacific Island economies are forecast to grow 2.2% in 2026, half a percentage point below the previous forecast. The drag is energy costs. These are small open economies with limited fiscal space, no semiconductor industry, no electronics assembly capacity, and no connection to the AI hardware supply chain. They are experiencing the energy cost consequence of the AI boom -- more electricity demand globally, higher fuel prices, higher import bills -- without any of the export revenue that the boom is generating elsewhere in the region. The World Bank notes that a sustained 50% rise in fuel prices could cut household income by 3 to 4% across the Pacific, with the poor and small businesses identified as the most exposed.

The adoption picture is similarly divided. Only 13% of jobs in the East Asia and Pacific region involve the kind of complex judgment tasks where AI is currently most useful as a productivity tool, against 39% in advanced economies. Between 13% and 17% of multinational subsidiaries in China and Thailand currently use AI -- roughly one third the rate seen in industrialised countries. The World Bank identifies digital connectivity gaps and skills shortages as the primary barriers. It recommends a focus on what it calls Small AI: adapting existing tools to local conditions rather than attempting to build large language models or costly computing infrastructure from scratch. The recommendation is sensible. It also acknowledges that the countries with the most to gain from AI adoption are precisely the countries least equipped to implement it.

The Corridor Test asks whether infrastructure and trade flows build productive capacity inside an economy or move value through it to external beneficiaries. Applied to East Asia's AI boom, the question becomes sharper: Malaysia, Vietnam and Thailand are assembling AI hardware -- computers, servers, routers -- for American and European technology companies. The assembly work keeps employment and some value-added inside those economies. But 69% of Thailand's AI-related exports and 60% of Vietnam's are assembled equipment. The design, the semiconductors, the software, the intellectual property, and the final commercial use are external. What stays in the country is the assembly margin, the employment, and the infrastructure investment that the foreign companies bring. That is a substantial gain. It is not the same gain as owning the technology.

The World Bank report notes that South Korea's two largest semiconductor companies accounted for 43% of the entire KOSPI stock market valuation at the end of April 2026. That is what it looks like when a country owns the technology rather than assembling it. South Korea designed and manufactures the chips. Malaysia and Vietnam assemble the products those chips go into. Both are on the AI wave. They are not on the same board.

The World Bank's October report is not pessimistic about the AI boom. Its growth forecasts for the region have been revised upward. But it names the tail risk with precision. A reversal in global AI activity would, in its words, remove a key pillar supporting growth in trade, investment and financial valuations across the region. The mechanism is straightforward: $800 billion of the planned $2.9 trillion in AI capital expenditure for 2025 to 2028 is expected to be financed through private credit. AI-related lending has risen from 18% of private credit activity to 34% in a single year. The Bank notes that a correction could mean investment ran ahead of realised demand -- not that the technology itself failed. The consequence for Malaysia, Vietnam and Thailand, whose export growth is now structurally dependent on that investment continuing, would be severe.