Korea's memory boom booms alongside a 25 basis point consumption drag from an aging population

The 25 basis points is multiplication before it is a forecast. The UN's 1.5 percentage points a year of dependency-ratio increase ^[4], times the top of the measured Korean elasticity of 17 basis points of consumption growth per point ^[5], gives 25.5 ^[7], and that is where Goldman's headline number lands ^[6]. Compound a quarter-point a year for ten years and the level of Korean consumption ends up roughly 2.5 per cent below where it would otherwise have been ^[8], which against one turn of the memory cycle is not much. The elasticity itself is the number worth stopping on: about three basis points for the average major economy in the sample against ten to seventeen for Korea ^[5], a multiple of between 3.3 and 5.7 ^[26], and gaps that wide come out of balance sheets, not moods.

Here is the balance sheet. More than 60 per cent of Korean household net worth is parked in non-financial assets, the highest share among the advanced economies Goldman studied, and household financial assets come to 100 per cent of 2024 GDP, the lowest in its sample ^[10]. Taiwan, carrying aging pressure of its own, runs net financial assets of about five times GDP and gets visibly stronger spending out of its retirees ^[13]. So the year's equity gain ^[2] has to reach consumption through a pipe in which fewer than one in four elderly households with retirement savings could actually fund their consumption from financial assets ^[11], and in which reverse mortgages reach 1.8 per cent of homeowners over 75 ^[12]. Koreans in their sixties keep 37 per cent of their income, saving harder than any other cohort ^[9]. That saving behavior, retirees out-saving workers, is exactly why the export surplus is not converting into retail sales.

Note what the capital is doing instead. The June 2026 public-private plan commits nearly $1.2 trillion, about two-thirds of GDP ^[15] (which implies a GDP near $1.8 trillion ^[16]), to fabs and data centres, while the current account is forecast above 10 per cent of GDP for 2026 ^[14], the national-accounts signature of a country producing more than it uses. Youth employment in AI-exposed sectors fell by 285,000 between June 2022 and June 2026, 94 per cent of it in IT services and programming ^[17], call it 268,000 posts ^[18]. Against that, the pro-natal offer of 20 million won per newborn ^[21] is small enough that a single chip bonus covers 27.6 of them ^[22].

This is probably wrong in one specific way, or rather, the more interesting version of wrong: financial assets at one times GDP is a stock, and stocks move. If a rerating pulls Korean household savings out of property and into securities, the drawdown channel Taiwan already has ^[13] starts to exist here, and the tests are dull and checkable: the household financial-asset-to-GDP ratio ^[10] and reverse-mortgage take-up moving off 1.8 per cent ^[12]. Or the semiconductor tax revenue earmarked for youth employment and housing ^[24] does its job, and the K closes from the bottom. Or memory pricing rolls over, the upper leg joins the lower one, and nobody asks about the dependency ratio again for a while.

The view, then. Own the earnings rather than the country: price Samsung Electronics and SK hynix ^[25] on high-bandwidth memory cash flow, and price Korean domestic demand on savers in their sixties who intend to leave the house to their children ^[9][12].