AI Props Up Global Trade, but Korea Risks a Jolt if Big Tech Spending Falters

AI-related products are rapidly taking over global merchandise trade growth. The dollar value of trade in AI-linked goods rose 42% in the first quarter from a year earlier, while all other goods managed only a 7% increase. Because AI investment helped the global economy withstand tariffs and an energy shock from the Middle East, any pullback in capital spending by U.S. tech giants would raise the risk of a simultaneous slowdown in semiconductors, power equipment, minerals and Asian exports.

AI Trade Up 42%, Non-AI Goods Up 7%

On September 19, the World Trade Organization said AI-related goods accounted for about 19% of total global merchandise trade in the first quarter. The category includes not only semiconductors, but also servers, telecommunications equipment, some industrial machinery and related components.

Total merchandise trade volume rose 3.2% in the first quarter from a year earlier, while the dollar value increased 11%. The divergence across products was pronounced. Trade in office and telecommunications equipment climbed 44%, and AI-related technology goods rose 42%. By contrast, chemicals fell 6%, steel dropped 5% and fuels declined 3%.

Financial Times cited WTO Director-General Ngozi Okonjo-Iweala as saying AI trade is masking the impact of tariffs and other disruptions surrounding global commerce. Global merchandise trade volume rose 4.6% last year, and the WTO estimates 42% of that increase came from AI-related goods. The surge in AI investment, in other words, goes well beyond a one-quarter phenomenon.

Recent leading indicators point the same way. The WTO’s goods trade barometer, released on September 9, stood at 102.0, above the long-term trend line of 100. The electronic components index came in at 104.9 and export orders at 103.5, while container shipping remained at 99.6. The message is that the recovery has not been broad-based across all goods. Electronic components and AI equipment are doing most of the lifting.

Asia Produces, North America Finances

The geographic concentration of AI trade is becoming clearer as well. Asia’s merchandise export volume rose 12.9% in the first quarter from a year earlier, while imports climbed 14.6%. That reflects growing intra-regional trade in AI-related parts moving among China, South Korea, Taiwan, Singapore and Thailand. Europe’s export volume fell 2.6% over the same period, though that decline was also shaped by a base effect after shipments of gold and pharmaceuticals were brought forward last year.

The FT reported that Asia handled more than 60% of global trade in AI-related goods last year. By contrast, 76% of venture investment in AI was concentrated in North America. The division of labor has become more entrenched: the U.S. provides design, capital and final cloud demand, while Asia, led by South Korea and Taiwan, supplies semiconductors and electronic components.

Okonjo-Iweala described a supply chain in which semiconductors are designed in one country, produced and packaged in another, and assembled into servers in a third. “Without trade, there is no computing,” she said. Singapore Minister for Digital Development and Information Josephine Teo said AI should become an opportunity for the many, not a privilege for the few. The point is that AI is both a growth engine and a sector where production and capital are concentrated in a handful of regions.

The trade rules underpinning the AI supply chain are older than the technology itself. The WTO’s Information Technology Agreement lowered tariffs on much of the equipment used in AI, while agreements on services, technical standards and intellectual property support the flow of data and technology. But those rules were written before large language models and AI accelerators emerged. If semiconductor export controls and local-production requirements expand, trade values may keep rising even as duplicate investment and supply-chain costs increase.

South Korea sits at the center of that global supply-chain structure. WTO data show the country’s nominal exports rose 38.4% in the first quarter from a year earlier, the fastest increase among the world’s five largest exporting economies. AI investment has created a circular structure that begins with U.S. capital spending, moves through South Korean memory chips, Taiwanese foundries, Asian assembly and logistics, and returns to U.S. data centers.

Big Tech Capital Spending Hits $730 Billion

The money supporting global trade is coming from capital spending by U.S. tech giants. LSEG data show projected capital expenditure this year for Microsoft, Alphabet, Amazon, Meta and Oracle rose to $730 billion in July from $485 billion in January. The companies do not break out AI spending separately, so the figure aggregates total capital expenditure. Still, a large share of investment in data centers, servers and networks is tied to AI demand.

Investment is rising faster than cash generation. On LSEG projections, capital spending at those five companies will exceed free cash flow in 2027. From 2025 through 2027, every additional $1 of operating cash flow would be matched by a $1.57 increase in capital spending. As David Russell, TradeStation’s head of market strategy, put it, companies exist to make money, not to spend it.

There is still no clear evidence that demand is weakening. Amazon raised its capital spending plan for this year by 10% to $220 billion. Second-quarter revenue at Amazon Web Services rose 37% to $42.2 billion, and backlog increased to $496 billion from $364 billion in the previous quarter. Chief Executive Officer Andy Jassy said AWS is growing explosively and that even $220 billion in spending would not be enough to meet all of this year’s demand.

Alphabet also raised its capital spending outlook for this year to $195 billion to $205 billion. Google Cloud revenue rose 82% in the second quarter to $24.8 billion. Chief Financial Officer Anat Ashkenazi said demand continues to outpace investment.

The strain is already showing up in cash flow. Amazon’s trailing 12-month free cash flow swung to a $7.6 billion deficit from a positive $18.2 billion a year earlier. Alphabet also reported negative free cash flow of $5.9 billion in the second quarter. Thomas Monteiro, an analyst at Investing.com, said capital once again carries a real cost and the margin for error is shrinking every quarter.

Bridgewater estimates Alphabet, Amazon, Meta and Microsoft will spend $650 billion on AI infrastructure this year, up 59% from last year’s $410 billion. That estimate covers a different universe from LSEG’s projection for total capital spending by five companies. Greg Jensen, Bridgewater’s co-chief investment officer, said competition in AI investment has entered a more dangerous phase. If expected returns weaken while dependence on external capital rises, investment and financial markets could both come under pressure.

Power and Minerals Are Part of the Same Cycle

AI trade does not end with semiconductors. The International Energy Agency projects global electricity consumption by data centers will almost double to 950 terawatt-hours by 2030 from 485 terawatt-hours last year. Power use by dedicated AI data centers will triple over the same period. By 2030, data centers will account for about 3% of global electricity demand.

Server power use is rising rapidly as well. The IEA said the power density of AI servers increased elevenfold from 2020 to 2025 and will quadruple again by 2027. By 2027, one server rack in an advanced data center could require as much electricity at peak load as 65 households. That is why transformers, power semiconductors, wires, batteries and transmission grids are being drawn into the AI supply chain.

IEA Executive Director Fatih Birol said electricity demand is growing three times faster than overall energy demand, and 61% of global energy investment this year has gone to the power sector. He also warned that grid constraints and high electricity prices could slow the pace of data-center expansion.

Mineral demand is tied to the same trend. Trade in ores and minerals rose 27% in the first quarter, while prices for metals and minerals excluding gold and silver increased 32%. Not all of that can be attributed to AI demand. Gold and copper prices, along with supply shocks from the Middle East, also played a role. Even so, investment in servers, transmission grids and cooling equipment is clearly helping drive trade in copper and electrical equipment.

Asia Would Be Hit First if Investment Slows

The WTO forecasts global merchandise trade volume will rise 1.9% this year under its baseline scenario. If AI investment remains strong, that growth rate could be 0.5 percentage point higher. If energy prices stay elevated because of war in the Middle East, it could be 0.5 point lower. In the WTO’s outlook, the lift from AI investment and the drag from energy shocks are roughly the same size.

If AI capital spending slows, the effects would spread in stages. A cut in data-center budgets by big tech companies would first hit orders for AI accelerators and high-bandwidth memory, or HBM. That would then reduce foundry, packaging and server assembly activity, along with intra-Asian trade in components. The fallout would spread further to air cargo, power equipment, copper orders and financing for data centers. The channels through which AI lifted global trade would begin working in reverse.

Okonjo-Iweala warned it would be difficult to sustain the current rise in trade if AI investment proves to be a bubble. The IEA also said data-center investment has grown beyond what corporate balance sheets can easily absorb, making future construction more sensitive to capital-market sentiment and expected returns.

There is little basis, however, for treating a sharp downturn as the baseline case right now. According to the Semiconductor Industry Association, global semiconductor sales reached $146.8 billion in July, up 135.1% from a year earlier. SIA President and CEO John Neuffer said the industry surpassed its previous annual global sales record in just seven months. Monthly sales have now increased for 17 straight months.

For South Korea, both opportunity and concentration risk are rising. Ministry of Trade, Industry and Energy data show exports in August rose 68.7% from a year earlier to $98.25 billion. Semiconductor exports jumped 209% to $46.65 billion, meaning a single product accounted for 47.5% of total exports.

Exports excluding semiconductors also rose about 20%. Even so, semiconductors accounted for about 79% of the total increase in exports. The export rebound has broadened, but the incremental gains have become even more concentrated in chips.

Kim Ju-wan, Hankyung.com reporter, kjwan@hankyung.com