South Korea's August chip exports hit $46.65B, nearly triple a year earlier
Semiconductors were 47.5% of the country's total exports. A national economy is now indexed to hyperscaler capital expenditure.
South Korea's chip exports reached $46.65 billion in August, nearly triple the figure a year earlier and 47.5 percent of the country's total exports, according to CNBC. The growth is driven by hyperscaler AI spending, primarily on memory from Samsung and SK Hynix.
Almost half of a developed economy's exports now come from one product category, sold to roughly a dozen customers, all of whom are executing capital expenditure programmes that depend on AI demand continuing to compound.
Why memory rather than logic
The accelerator gets the attention; high-bandwidth memory is the actual constraint, and it is where Korea's position is.
HBM is manufactured by stacking DRAM dies vertically with through-silicon vias, and the yield, thermal and packaging challenges make it far harder than conventional memory. Samsung and SK Hynix, with Micron, are effectively the world's supply. Every Nvidia accelerator, every Google TPU, every AMD MI-series part and every hyperscaler's internal silicon needs it.
That is why the tripling is possible: AI accelerators consume far more memory per unit than anything before them, and the memory is priced accordingly.
It is also why the export control regime reaches further than accelerators. A Chinese programme built on domestic Huawei silicon — DeepSeek is planning a cluster of more than 160,000 accelerators at a one-gigawatt site in Inner Mongolia — still needs HBM, and that is a chokepoint two Korean companies and one American one control.
The concentration risk runs both ways
For Korea, 47.5 percent of exports in one category is an extraordinary dependency. The demand comes from a handful of firms whose spending is set by AI capital expenditure plans, and those plans are underwritten by valuations rather than by current cash flows.
The warning signs in the same fortnight's reporting are not subtle. Big Tech booked more than $160 billion in second-quarter "other income" from AI stakes, with Alphabet alone at $97.983 billion — largely unrealised gains, roughly four times its $24.8 billion of quarterly Google Cloud revenue. Nvidia is investing in the compute providers that buy its chips, leasing the buildings they sit in, and forecasting 70 percent growth. SB Energy granted OpenAI $5.5 billion in warrants ahead of its own IPO.
If that structure repriced, Korean export figures would show it within a quarter.
Seoul is spending into it rather than hedging
The policy response has been to double down. South Korea has committed $919 billion to a sovereign AI programme targeting 8.4GW of capacity by 2029 and 18.4GW by 2035, coordinating power, compute and domestic model development across SK Group, GS Group and Naver.
Samsung also led Mistral's €3 billion Series D this month at a valuation above €21 billion, which President Macron framed as France and South Korea building "a third way in AI."
That is a strategy of moving up the stack: from selling memory into other people's AI programmes to having a domestic one, and to holding equity in European model developers who will buy Korean memory. It is a reasonable hedge against being permanently a supplier.
It does not reduce the near-term exposure, and it is financed by the same revenue stream it is hedging.
Runs the newsroom. Rename this profile in the studio to your own byline.
Related
Every weekday, the AI stories that moved money or shipped code.
No cross-posting, unsubscribe anytime. See all newsletters