The AI capex boom is spilling over strongly into Asian tech-exporting economies

Overview

The report analyzes the spillover effects of rising U.S. and global AI-related capital expenditure on Asian economies. The core logic is that most hardware spending in data center investment is concentrated in AI servers, storage, and network equipment, and these products are mainly produced, assembled, or supplied with key components by Asian supply chains. Goldman Sachs expects global AI-related capex could exceed $1 trillion in 2026, with hardware spending above $600 billion; the consensus expectation for 2026 capex from the five major US hyperscalers is around $750 billion.

Core views

The core view is that the AI investment boom has already and will continue to significantly lift growth in Asian tech-exporting economies, with Taiwan as the most AI-leveraged economy in Asia and globally; Korea benefits from HBM and NAND memory; Malaysia, Singapore, Vietnam, and Thailand also see material gains in packaging and testing, assembly, chip manufacturing, or data center investment. By contrast, although China has a huge electronics manufacturing scale, direct growth acceleration is relatively muted because major nodes in the advanced-process semiconductor core supply chain are largely located outside China. The report also argues that strong technology exports will improve current accounts in economies like Taiwan and Korea, and AI export tailwinds can offset part of the shock even if energy prices rise.

Analysis framework

The report uses multiple methods to estimate the contribution of AI-related tech exports to growth: first, using industry-level GDP or gross value-added data to see the contribution of manufacturing sub-sectors such as electronics, computers, and optical products to real GDP growth; second, using tech export data and OECD 2022 trade value-added data to estimate local value added in technology exports and convert it to real terms using export price indexes; third, predicting 2026 contributions through the regression relationship between US hyperscaler capex and each economy’s technology exports; in addition, it also uses Taiwan and Korea’s directly released nominal and real technology value-added data, as well as Korea chip shipment data, for cross-checking.

Methodology notes

• Macro growth decompositionIndustry GDP contribution method

  Estimate the impact of the technology sector on growth by using the contribution of manufacturing sub-sectors such as electronics, computers, and optical products to real GDP growth.

  The method comes directly from national statistical accounting, suitable for economies like Taiwan and Korea that have disaggregated industry data; the drawback is that the industry coverage is often broader than AI-related equipment, and official GDP data may be revised in the future.

• Trade value-added analysisTech export local value-added method

  Estimate the contribution of local value added in technology exports to GDP growth using tech export amount, OECD TIVA local value-added share, and export price indexes.

  This method is more timely and intuitively linked to external capex, but it relies on extrapolating 2022 value-added shares, price-index choices, and supply-chain structure assumptions, so estimation errors can be large.

• Forecast modelHyperscaler capex regression method

  Predict 2026 export contributions using each economy’s sensitivity of technology exports to US hyperscaler capex.

  This method is used to convert the 2026 consensus expectation for US hyperscaler capex into regional export growth forecasts; the report notes the sample is only about 13 quarters, so results should be treated as approximate.

Asset mapping & comparison

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

• Taiwan macro economy and TWD

  The most direct beneficiary of AI hardware capex and technology exports

  Strengths

  Advanced logic chips and AI server manufacturing are at the core of the supply chain, with estimated technology-export contribution in 2025 as high as 5.9 percentage points and still forecast at 4.5 percentage points in 2026, and an extremely high current account surplus.

  Weaknesses

  Growth is highly sensitive to the technology-export and AI capex cycle, potentially leading to K-shaped recovery and higher volatility.

  Comparison

  Compared with other Asian economies, Taiwan is the most AI-leveraged economy in the report.

  Risks

  AI capex slowdown, export price index errors, supply-chain migration, energy price shocks, and official data revisions.

• Korea macro economy, MSCI Korea, and KRW

  Benefits from demand for high-end memory like HBM and NAND

  Strengths

  SK Hynix and Samsung are core in high-bandwidth memory and data-center storage, with the 2026 technology-export contribution expected to be about 1.0 percentage point and notable current account improvement.

  Weaknesses

  Non-tech exports and consumption are relatively weak, with clear K-shaped growth features.

  Comparison

  Benefit is less than Taiwan, but greater than most non-core tech-exporting economies.

  Risks

  Memory-cycle volatility, cooling AI server demand, energy import costs, and weak domestic demand.

• Malaysia macro economy and MYR

  Benefits through semiconductor packaging and testing, power devices, and data center investment

  Strengths

  Has mature OSAT and packaging-testing capabilities, with a meaningful share of global semiconductor ATP, expected tech contribution around 2.0 percentage points in 2026, and relatively large data-center investment versus GDP scale.

  Weaknesses

  Compared with advanced logic chip and AI server core links, added value and pricing power are weaker.

  Comparison

  Among Southeast Asia, the benefits stand out, and the report is relatively bullish on MYR.

  Risks

  Margin pressure in lower-end supply-chain stages, uncertainty around foreign-capital project realization, and volatility in energy and global demand.

• Singapore macro economy

  Benefits from equipment production, chip manufacturing, packaging and testing, and electronics exports

  Strengths

  Has a meaningful share of global chip and semiconductor equipment production, with electronics exports showing a clear recent increase.

  Weaknesses

  Relative to its economic size, data-center investment pull may be smaller than in Malaysia and Thailand.

  Comparison

  High tech exposure, but growth contribution is estimated lower than Taiwan, Malaysia, and Vietnam.

  Risks

  Global electronics-cycle swings, concentrated external demand, and potential pressure on services exports.

• Vietnam macro economy

  Benefits from electronics assembly and emerging packaging and testing links

  Strengths

  Estimated tech-export contribution is relatively high, with a 2026 forecast around 3.4 percentage points.

  Weaknesses

  The report indicates most technology exports are not AI-related, so direct AI leverage could be overstated.

  Comparison

  Contribution figures are high, but AI purity is lower than Taiwan and Korea.

  Risks

  Non-AI electronics cycle, uncertainty around foreign manufacturing relocation, and local value-added share estimation error.

• Thailand macro economy

  Indirectly benefits through backend packaging and testing and data center investment

  Strengths

  Recently approved large-scale data center projects, and 2026 data center investment contribution to growth may exceed 1 percentage point.

  Weaknesses

  Tech exports contribute little to total growth, and energy import pressure is large.

  Comparison

  AI export tailwind is weaker than in Taiwan, Korea, Malaysia, and Vietnam.

  Risks

  Energy prices, project execution pace, and volatility in tourism and domestic demand.

• China macro economy and CNY

  Benefits through lower-tier components, partial domestic AI investment, and participation in regional supply chains

  Strengths

  Very large electronics manufacturing scale, supplying many passive components, materials, and lower-tier parts; the report is relatively constructive on CNY.

  Weaknesses

  Many key nodes in advanced-process semiconductor core supply chain are outside China and are affected by export controls, so direct AI supply-chain pull is relatively modest.

  Comparison

  Overall tech manufacturing scale is large, but AI growth sensitivity is lower than Taiwan and Korea.

  Risks

  Export controls, geopolitics, traditional electronics demand, and domestic AI capex not meeting expectations.

• Japan macro economy

  Benefits through upstream semiconductor equipment and materials

  Strengths

  Has key supply capacity in wafers, photoresist, and semiconductor manufacturing equipment.

  Weaknesses

  Tech exports and sector GVA contributions to overall GDP growth are relatively low.

  Comparison

  More of an upstream materials-equipment role, with smaller macro growth sensitivity than Taiwan, Korea, and some Southeast Asian economies.

  Risks

  Global semiconductor equipment cycle, yen volatility, and slowing external demand.

• MSCI Taiwan and MSCI Korea versus MSCI India, Philippines, Thailand

  The report cites market and portfolio-strategy commentary favoring North Asia markets relative to South Asia and some Southeast Asian markets

  Strengths

  North Asia’s growth revisions are supported by AI exports, and Taiwan and Korea forecasts are above consensus.

  Weaknesses

  The market may have already priced in part of the AI tailwind.

  Comparison

  The report mentions a recent preference to be long MSCI Korea and Taiwan, and short MSCI India, Philippines, and Thailand.

  Risks

  Downward revisions in AI capex expectations, changes in energy shocks, valuation crowding, and FX volatility.

Key data

• Share of data center capex in hardwareabout 60%-70%Mainly AI servers, storage, and network equipment, almost all of which are manufactured in Asia.
• Global AI capex in 2026likely over $1 trillionThe report says global AI-related investment is materially above the measure from the five major US hyperscalers.
• 2026 capex of the five major US hyperscalersabout $750 billionBased on post-Q1 earnings market consensus expectation.
• Expected server revenue in 2026about $650 billionGoldman Sachs Technology Research team expects most of it to be AI-related.
• Incremental Asia-related electronics exports to the USover $200 billion annualizedFrom before ChatGPT’s launch in 2022 to early 2026.
• Taiwan 2025 growth contribution from technology-related exports5.9 percentage pointsEstimated value in the report, reflecting Taiwan’s central role in advanced logic chip and AI server manufacturing.
• Taiwan 2026 tech-related export growth contribution forecast4.5 percentage pointsAssuming AI investment continues to grow rapidly, though growth decelerates.
• Korea 2025 and 2026 contributionabout 0.8 percentage points, 1.0 percentage pointsKorea is an important supplier of high-end memory, HBM, and NAND.
• Malaysia 2026 technology sector contribution forecastabout 2.0 percentage pointsSupported by packaging and testing, power devices, and data center investment.
• Vietnam 2026 technology sector contribution forecastabout 3.4 percentage pointsThe report notes most of Vietnam’s tech exports are not AI-related.
• Taiwan 2025 current account surplusclose to 20% of GDPThe report says the current tracked level in 2026 is above this number.
• Korea 2025 current account surplus6.6% of GDPThe report expects the surplus to move into double digits in 2026 even with higher oil prices.

Impact & implications

The AI capex cycle will strengthen divergence in growth, current accounts, and market performance across Asia. North Asian tech-exporting economies, especially Taiwan and Korea, may continue to benefit from AI exports in growth forecasts, export prices, and current account balances, supporting related equity markets and some currency performance; South Asia and some Southeast Asian energy-importing economies are more vulnerable to oil-price shocks, and AI export tailwinds may not fully offset energy headwinds. The report also notes that AI-driven growth may be concentrated in capital-intensive manufacturing and may not simultaneously translate into strong jobs, wages, consumption, or inflation pressure.

Risks

• AI-related capex growth slows or consensus expectations for hyperscaler investment are revised lower.
• Growth becomes overly concentrated in tech manufacturing, resulting in K-shaped recovery, limited labor spillovers, and insufficient domestic demand improvement.
• High energy prices weaken growth and current accounts in energy-importing economies and partially offset AI export tailwinds.
• There are estimation errors and future revision risks in technology exports, price indexes, local value-added shares, and GDP industry data.
• Supply chains continue to migrate because of tariffs, geopolitics, and export controls, changing the benefit distribution across economies.
• Public data coverage of data center investment is inconsistent; project approvals do not equal actual implementation.
• Service exports may face negative impacts from the AI investment boom or changes in resource allocation; the report says this will be researched separately.

What to watch

• Whether 2026 capex from the five major US hyperscalers is close to or above the $750 billion consensus expectation.
• Whether global AI capex stays above the $1 trillion scale and hardware spending remains above $600 billion.
• Three-month annualized trends in Taiwan exports to the US, US net imports of AI hardware, and Asian technology exports.
• Whether Taiwan and Korea tech manufacturing GVA, export prices, and real export volumes continue to stay above expectations.
• Order and pricing changes in HBM, NAND, complete AI servers, liquid cooling, power supplies, and packaging and testing.
• Actual starts, imported equipment, and GDP contribution of data center projects in Malaysia and Thailand.
• The impact of energy prices on current accounts and growth forecasts of energy-importing economies such as India, Thailand, Indonesia, and the Philippines.
• Whether regional current-account divergence in North Asia continues to support CNY, TWD, and MYR.
• Whether North Asian equity index performance versus India, Philippines, and Thailand can continue.