The AI boom is unlikely to translate quickly into consumption growth in South Korea, while population aging poses a more persistent headwind

Overview

The report examines whether the AI boom can transmit beyond South Korean chip exports and capital expenditure to household consumption. It concludes that exchange-rate and fiscal channels may provide support after some time, but rapid population aging, continued saving by elderly households, and insufficient asset liquidity will make this transmission slow and uneven.

Core views

The AI boom is clearly driving South Korean exports and capital expenditure, particularly chip exports and related investment, but retail sales and broader private consumption remain weak, creating a “K-shaped” cycle of strong exports and investment but weak domestic demand. Taiwan, China, exhibits a similar divergence, although to a lesser degree, because its broader technology ecosystem—including AI servers, components, and advanced logic chips—has benefited directly from AI demand for longer and across a wider scope. Consequently, considerable uncertainty remains over whether and when AI-driven growth will benefit other sectors of the South Korean economy. Transmission of the AI boom to consumption mainly depends on channels such as the exchange rate, fiscal revenue, and household income. The depreciation of the Korean won early in the year and the fiscal authorities' decision to save part of the revenue windfall may both limit near-term consumption gains. If the won ultimately stabilizes in the second half, it could delay or offset oil-price-driven inflation because exchange-rate pass-through to inflation is relatively rapid in South Korea. The report uses Taiwan, China, as a comparison: despite real GDP growth exceeding 10%, more than five times its potential growth rate, inflation remained below 2%, demonstrating that strong technology-sector growth does not necessarily generate immediate, broad-based demand and inflationary pressures. Demographics represent a more critical and persistent constraint on South Korean consumption. The baby boomer generation born after the Korean War is entering retirement, while the fertility rate remains below 1. The working-age population has been declining since 2017, while the elderly population is increasing rapidly. The United Nations projects that South Korea's total dependency ratio will rise by 1.5 percentage points annually over the next decade, the fastest among 70 medium-sized and large economies and even above Japan's peak increase of 1.3 percentage points per year during 2000–2015. South Korea's current dependency ratio remains below that of most major economies, but the speed of increase means this advantage will disappear rapidly. Goldman Sachs uses panel regressions to assess the impact of aging on consumption. After using real GDP growth as a proxy for real disposable income growth and controlling for time fixed effects, it finds that, across major economies, each 1 percentage point increase in the dependency ratio reduces real private consumption growth by an average of 3 basis points. On a per capita basis, the decline is slightly larger at 4–5 basis points. South Korea's sensitivity is significantly higher: country-specific regressions show that each 1 percentage point increase reduces consumption growth by 17 basis points. Under an alternative specification using an unrestricted lag structure, the decline is still 10 basis points. The report therefore presents 10–17 basis points as a reasonable range. Assuming that the dependency ratio rises by 1.5 percentage points annually over the next decade and consumption sensitivity is 17 basis points, South Korean consumption growth would face an annual drag of approximately 25 basis points. Goldman Sachs' top-down simulation indicates that consumption growth will continue to slow over the next two decades, subsequently turn negative, and increasingly lag real GDP growth. The simulation still uses a relatively optimistic assumption that AI-driven productivity gains and structural reforms will enable South Korea to sustain long-term GDP growth of 2% over the next two decades. This assumption is well above the sub-0.5% forecast from a South Korean government think tank and also above Goldman Sachs' estimate of approximately 1% through 2075 based on standard growth accounting. Even under this stronger growth assumption, demographics still depress consumption performance. The relatively high propensity to save among South Korea's elderly population explains why the consumption drag is greater than in other economies. Household surveys show that the savings rate of people aged 60–69 is higher than that of other age groups. The savings rate of those aged over 70 remains similar to that of people aged 40–49, whereas residents in many economies typically reduce saving significantly after retirement to support consumption. After briefly declining during the global financial crisis, South Korea's elderly savings rate has continued to rise, contrasting with recent post-pandemic declines in savings rates in Japan and Taiwan, China. A bottom-up simulation based on age composition further supports this assessment. If the savings trends of each age group over the past decade persist, rapid aging will increase the surplus of household cash income over cash expenditure by 30 basis points annually over the next decade. After incorporating social transfers in kind and owner-occupied housing costs and converting the result to a national accounts basis, this equates to a 15-basis-point annual increase in the household savings rate. The shift in population weights toward high-saving age groups means that even if income is supported by the AI industry, a large share of additional income may be saved rather than consumed. Household balance sheets reinforce this mechanism. In 2024, South Korean households' net financial assets were approximately 100% of GDP, the lowest among the major advanced economies in the report's sample, while those of Taiwan, China, were close to five times GDP. The composition of gross financial assets in the two economies is broadly similar; the main difference is that South Korean households have more debt. Household debt is equivalent to approximately 40% of net household assets, or 90% of GDP. The larger net financial buffer held by households in Taiwan, China, may be an important reason why aging there has not produced an equivalent drag on consumption. More than 60% of South Korean household assets consist of real estate and other nonfinancial assets, the highest share in the sample. These assets are difficult to monetize and are typically financed by bank loans, which both limits retired households' ability to smooth consumption and creates potential solvency risks if housing prices undergo a substantial correction. Surveys show that when income declines, most South Korean households cut consumption or increase earned income rather than sell assets. Even among elderly households that have accumulated retirement savings, fewer than one-quarter can meet their consumption needs using financial assets alone. Australia has a similar share of nonfinancial assets, but its population is younger, with a median age of 38 compared with 46 in South Korea. The use of housing wealth release tools remains limited. The South Korean government has promoted reverse mortgages to alleviate liquidity constraints among retired households and reduce excessive saving, but the product covers only 1.8% of homeowners aged over 75. Strong bequest motives are one reason for the low adoption rate, while the underdevelopment of trust services also limits the systematic management of retirement assets. From a policy perspective, the report argues that South Korea needs more time and more effective transmission mechanisms to convert AI-driven growth into sustained consumption. Stabilization of the Korean won would help ease inflationary pressure and limit the risk of a further increase in real interest rates under inflation targeting. Fiscal authorities should save part of the revenue windfall while redirecting funds toward productivity-enhancing investment. Capital-market reforms and wider use of reverse mortgages and other home-equity tools could alleviate cash-flow constraints among retired households. Monetary policy may still need to remain restrictive, but following the precautionary rate hike in August, further tightening is expected to be cautious given the high uncertainty surrounding the timing and scale of the transmission of AI dividends, requiring close coordination among fiscal, monetary, and financial policies.

Analysis framework

The report first compares the differing effects of the AI boom on South Korean exports, investment, and consumption, and then examines short-term transmission through exchange-rate and fiscal channels. It subsequently estimates the impact of changes in the dependency ratio on consumption growth using cross-country panel regressions and country-specific regressions for South Korea, testing the results with an alternative lag specification. Finally, it combines age-group savings rates in a bottom-up simulation, conducts a top-down consumption simulation under long-term growth assumptions, and uses cross-country household balance sheets, housing-asset liquidity, and the use of policy tools to explain why South Korea is more sensitive to aging.

Methodology notes

• Macroeconomic framework

  Cross-country panel regressions and country-specific regressions

  The report controls for real GDP growth and common time factors to estimate the average impact of a rising dependency ratio on private consumption growth, and then separately estimates South Korea's sensitivity to identify how it differs from other economies.

• Macroeconomic framework

  Robustness test using an unrestricted lag structure

  The report allows demographic changes to affect consumption over different horizons and uses an alternative model to test its main conclusion. The resulting decline of 10 basis points is smaller than the baseline estimate of 17 basis points, but the direction and economic significance remain unchanged.

• Macroeconomic framework

  Bottom-up simulation of age-group savings rates

  This method combines the saving behavior of each age group with changes in future population weights to estimate how aging raises the overall household savings rate and depresses consumption.

• Macroeconomic framework

  Top-down consumption simulation under demographic and growth assumptions

  The report applies the dependency-ratio path and consumption sensitivity to a long-term growth scenario to assess future consumption growth and its performance relative to GDP growth, using standard growth-accounting forecasts as a comparison benchmark for its long-term assumptions.

Key data

• Increase in South Korea's total dependency ratio over the next decade1.5 percentage points annuallyUnited Nations forecast; the fastest among 70 medium-sized and large economies
• Historical peak increase in Japan's dependency ratio1.3 percentage points annuallyDuring 2000–2015, below the projected pace for South Korea over the next decade
• Average cross-country consumption sensitivityDecline of 3 basis pointsChange in real private consumption growth corresponding to each 1 percentage point increase in the dependency ratio
• Cross-country per capita consumption sensitivityDecline of 4–5 basis pointsPer capita estimate for each 1 percentage point increase in the dependency ratio
• South Korean consumption sensitivityDecline of 17 basis pointsAnnual impact of each 1 percentage point increase in the dependency ratio in country-specific regressions
• South Korean consumption sensitivity under the alternative modelDecline of 10 basis pointsEstimate using an unrestricted lag structure
• Drag on South Korean consumption growth over the next decadeApproximately 25 basis points annuallyCalculated based on a 1.5-percentage-point annual increase in the dependency ratio and sensitivity of 17 basis points
• Increase in household cash surplus30 basis points annuallyAssuming the age-group savings trends of the past decade persist
• Increase in the household savings rate15 basis points annuallyNational accounts basis after incorporating social transfers in kind and owner-occupied housing costs
• South Korean long-term GDP growth assumption2%Top-down simulation assumption for the next two decades, supported by AI-driven productivity gains and structural reforms
• Other long-term growth forecasts for South KoreaBelow 0.5%; approximately 1%Respectively, a government think tank forecast and Goldman Sachs' estimate through 2075 based on standard growth accounting
• South Korean household net financial assetsApproximately 100% of GDPIn 2024, the lowest among major advanced economies in the sample
• Taiwan, China, household net financial assetsClose to five times GDPSignificantly above South Korea, providing a larger financial buffer
• South Korean household debtApproximately 40% of net household assets, or 90% of GDPIn 2024, significantly higher than that of households in Taiwan, China
• Share of nonfinancial assets in South KoreaMore than 60%Primarily real estate; the highest among the advanced economies in the report's sample
• Reverse mortgage coverage rate1.8%Refers only to homeowners aged over 75
• Median age comparisonSouth Korea: 46; Australia: 38The two countries have similar household shares of nonfinancial assets, but South Korea's population is significantly older
• Taiwan, China, growth and inflation comparisonReal GDP growth exceeded 10%, while inflation was below 2%GDP growth exceeded potential growth by more than five times but did not generate the strong inflation typically expected

Impact & implications

The report argues that growth in South Korea's AI industry will not automatically or immediately translate into a broad-based consumption boom. Demographics and household balance sheets may cause consumption to lag GDP over the long term. Policy priorities therefore include not only sustaining technology-sector growth but also stabilizing the exchange rate, improving the allocation of fiscal resources, raising productivity, and helping retired households convert illiquid assets such as housing into disposable cash flow.

Risks

• There is considerable uncertainty over the timing and scale of the AI boom's transmission to consumption through exchange-rate, fiscal, and household-income channels.
• Excess manufacturing capacity in the region may continue to weigh on non-chip exports from South Korea and other regional export-oriented economies.
• A fertility rate below 1, a shrinking working-age population, and a growing elderly population may cause the dependency ratio to rise faster than the consumption system can adjust.
• Depreciation of the Korean won may intensify imported inflation and increase the risk of rising real interest rates and further monetary policy tightening.
• A large volume of illiquid housing assets financed by bank loans may create household solvency risks if housing prices undergo a substantial correction.

What to watch

• Monitor whether the Korean won stabilizes in the second half and how exchange-rate changes transmit to oil-related inflation and real interest rates.
• Monitor when and to what extent growth in AI exports and capital expenditure transmits to household income, retail sales, and private consumption.
• Monitor how fiscal authorities handle AI-related revenue windfalls and how much is redirected toward productivity-enhancing investment.
• Monitor whether South Korea's total dependency ratio, elderly savings rate, and overall household savings rate rise along the paths projected in the report.
• Monitor the coverage rate of reverse mortgages and the use of trust services and other home-equity release tools.
• Monitor whether the Bank of Korea tightens further following its precautionary rate hike in August, as well as the coordination of monetary, fiscal, and financial policies.