SK hynix: record ₩104.6T revenue, 80% OPM, 'sold out through 2029' and a ₩100T ($75B) shareholder return — the shortage cycle becomes contract fact; same night, two CapEx verdicts: Microsoft cuts FY26 capex to ~$175B (Azure +48%) and jumps 4.3% after hours while Alphabet holds 2026 capex and promises 'significant increase' in 2027 on a revenue miss, sinking 10%; Anthropic's public S-1 puts an $808B commitment book and data-center moratoriums on the record

1. SK hynix writes the shortage cycle into contract: ₩104.6T revenue (+328%), 80% operating margin, CFO says "sold out through 2029, constrained through 2030" — then answers "where do the profits go" with a ₩100T ($75B) shareholder return

What happened. At its 10/28 earnings call: Q3 revenue ₩104.62T ($77.99B, +328%), operating profit ₩83.33T (+632%), net profit ₩39.78T — the company's first-ever quarter above ₩100T in revenue, with an operating margin of 80%. CFO Kim Woo-hyun, verbatim: DRAM and HBM capacity is "effectively sold out through 2029"; "we expect it to be difficult to fully meet customers' volume requirements in the near term… a situation where supply cannot keep up with demand is expected to persist for some time, a constraint projected to continue through 2030"; LTAs plus contracts still under discussion cover the next three years of expected output (Motley Fool call relay). Capital allocation landed bigger and more conventional than the rumor: a ₩100T ($75B) shareholder-return plan for this year — at least ₩30T in cash dividends plus more than ₩60T in buybacks and cancellations (including the completed ₩40T buyback pending cancellation); the Q3 board also approved ₩14T in quarterly dividends (₩20,000/share) and a fresh ₩20T buyback. The company's own words: "our shares are undervalued… repurchasing and canceling shares is an effective way to increase value per share."

Why it matters.①"Sold out through 2029" upgrades the shortage cycle from spot-market narrative to contract fact: Micron's August framing ("2027 even tighter than 2026") is now certified by the Korean incumbent and extended to 2030 — the duration of pricing power is measured in years;②HBM's cannibalization of conventional capacity is the multiplier (MU's framing: the same bits of HBM4E consume ~4x standard DRAM capacity), so "AI tight / conventional also tight" now has official confirmation;③The ₩100T return is the shortage cycle's first systematic monetization: not a capex arms-race one-way street — buy-and-cancel directly shrinks the share count. Two caliber notes: last week's rumored Kioxia stake sale + ₩16T special dividend was NOT confirmed (replaced by a larger, more conventional package); and ₩10T of the new ₩20T buyback is employee compensation excluded from the ₩100T return tally — the effective market-facing number is ~₩90T.

Investment implications. Memory's business model now explicitly splits into "contract annuity (LTA-locked 2027-29) + spot option (2030+)" — the sell side is already repricing (Seeking Alpha: "sold out through 2029, YE27 5x P/E doesn't make sense"). The market stamped it in sync: SKH Seoul +10.89% to ₩1,487,000, SKHY ADR +13.30% to $164.70, KOSPI +6.63% to 6,012.44; MU +5.86%, SNDK +7.66%, WDC +4.08% across the board. Falsified by: LTA renegotiations, spot weakness in 2028, or 2027 supply arriving materially early.

HIGH IMPACT Confidence: HIGH (company financials and call first-hand; Reuters/CNBC/Korea Times multi-source) Affected: 000660 KS · SKHY US · MU US · 005930 KS · SNDK US