Summary
Seoul Economic Daily quantifies where the AI memory windfall’s cost actually lands — and it is not (mechanically) on customers. SK Hynix faces ~100 trillion won in combined shareholder returns + employee performance bonuses in 2026, growing ~+450% YoY while CAPEX + R&D — the spending “directly tied to future competitiveness” — grows only +20–30%. This is the cleanest sourced statement of the AI Cost Incidence question at the chip layer: the profit-share windfall is absorbed by shareholders and reinvestment headroom, raising a competitiveness/crowd-out concern, rather than being passed through to memory buyers. Directly relevant to the June 5 flagship’s hyperscaler-cost question (GAP 4).
Key Points
- Combined reward burden ~100 trillion won (2026):
- Shareholder returns: 45–90T won (dividends, buybacks, retirements) — up to 7× last year’s 13T won; 50% of free cash flow allocated to shareholder returns.
- Employee performance bonuses: ~23T won — calculated at 10% of operating profit, avg ~600M won/employee, nearly 5× last year’s 4.7T won.
- Reward growth vastly outpaces investment growth:
Category Last year This year Change CAPEX + R&D 36.64T won ~50T won +20–30% Shareholder + Employee rewards ~18T won ~100T won +~450% - The competitiveness concern (verbatim framing): “the pace of rising expenses far exceeds the growth in capital expenditures (CAPEX) and research and development (R&D) spending, which are directly tied to future competitiveness.”
- Four big expense categories (shareholder returns, bonuses, CAPEX/R&D, taxes) projected to “combined approach around 200 trillion won” in 2026.
- U.S. ADR-listing tension: “Questions are being raised over whether it is appropriate to burden the company with large performance bonus payments while pursuing an additional overseas listing.”
- Company counter: bonuses are “a future investment to prevent talent drain and boost productivity.”
Newsletter Angles
- This reframes GAP 4 honestly. The flagship’s tempting claim — that the labor windfall flows into chip prices hyperscalers pay — is not what the cost-structure evidence shows. The windfall is a profit-share; its incidence lands on shareholders and reinvestment capacity. Pass-through to buyers would require the chokepoint to use its pricing power to rebuild margin — a strategic choice, not an automatic cost flow.
- The crowd-out is the real second-order story: rewards +450% vs. CAPEX/R&D +20–30% means the windfall-sharing fight competes with the reinvestment that sustains the chokepoint. That’s a sharper, better-sourced angle than a forced pass-through claim.
- Pairs with AI Cost Incidence’s electricity-side evidence to show incidence operating at two layers: ratepayers absorb the grid cost; shareholders/reinvestment absorb the chip-labor cost.
Entities Mentioned
- SK Hynix — subject; the 100T won reward burden
- Samsung — parallel windfall-sharing case (10.5%-of-OP deal)
Concepts Mentioned
- AI Cost Incidence — chip-layer incidence (rewards vs. reinvestment)
- AI Windfall Sharing — the upside-distribution companion
- Chokepoint Control
Quotes
“The pace of rising expenses far exceeds the growth in capital expenditures (CAPEX) and research and development (R&D) spending, which are directly tied to future competitiveness.” — industry observers, via Seoul Economic Daily
Notes
Solid financial journalism with hard figures. About SK Hynix (not Samsung), but the cost-structure logic transfers directly to Samsung’s larger 10.5%-of-OP deal. The load-bearing point for the flagship: the windfall is a profit-share (a claim on realized OP), so incidence falls on shareholders/reinvestment, not (mechanically) on chip prices. WebFetch succeeded.