Samsung, SK Hynix reject KEPCO’s $19 billion power prepayment proposal, document shows

Samsung Electronics and SK Hynix rejected Korea Electric Power Corp.'s proposal for a combined 25 trillion won ($18.7 billion) upfront payment to support electricity infrastructure for planned semiconductor mega-clusters. The chipmakers questioned whether the advance funding was necessary, citing uncertainty about the long-term durability of semiconductor demand. The dispute could delay or complicate power-infrastructure investment needed for new Korean chip fabrication capacity.

Analysis

The dispute exposes an off-balance-sheet subsidy embedded in Korea’s semiconductor-cluster strategy: grid capex required for future fabs may be materially larger, and arrive earlier, than the electricity economics currently reflected in Samsung and SK Hynix expansion assumptions. For SK Hynix, the near-term benefit is preserving liquidity for HBM capacity, where returns are demonstrably higher than funding regulated transmission assets. The longer-term risk is not the payment itself but a delay in grid interconnection that pushes out incremental wafer starts and weakens Korea’s ability to capture AI-memory demand versus Micron and Taiwanese supply chains.

KEPCO is the cleaner negative read. If it must finance dedicated generation and transmission through additional debt rather than customer prepayments, the burden falls on a balance sheet already highly sensitive to fuel costs, tariff decisions and financing rates. A regulated recovery mechanism could eventually protect economics, but the likely interim outcome is lower free cash flow, a higher allowed-return debate, and renewed political pressure against household tariff increases; that is a 6-18 month multiple constraint rather than necessarily an immediate earnings event.

Consensus may overstate the direct hit to SK Hynix: a large upfront infrastructure contribution is economically equivalent to a lower future power tariff only if it is non-refundable and earns no contractual return. The key diligence item is whether authorities can mandate a cost-sharing formula or offer capacity priority/tariff concessions in exchange. Until that is clarified, this is principally a timing risk to Korean fab additions, not evidence of weakening AI-memory end demand.

Over the next 1-3 months, watch government announcements on grid-routing approvals, KEPCO’s capex and funding plan, and any revision to semiconductor-cluster construction schedules. Thesis is falsified for the bearish KEP view if the utility secures explicit state capital support, binding customer deposits, or a tariff framework that preserves returns; it is falsified for a SK Hynix supply-delay concern if grid milestones remain on schedule without incremental customer funding.

Key Decisions for Investors

• Maintain SK Hynix (000660 KS / SKHY) exposure rather than reduce on this headline; reassess only if disclosed cluster commissioning dates slip by more than 6-12 months or management raises non-HBM capex materially. Near-term liquidity preservation is favorable, while the main risk is deferred capacity rather than an immediate P&L charge.
• Avoid or underweight KEPCO (015760 KS / KEP) for the next 1-3 months pending its financing and capex disclosure. A relative trade—long SK Hynix versus short KEPCO—expresses the transfer of infrastructure-funding risk, but size modestly because government intervention can rapidly re-rate KEP.
• Use Micron (MU) as a hedge/watch-list beneficiary if Korean grid delays become schedule-specific: a 6-18 month delay in Korean HBM/DRAM additions would tighten high-end memory supply and improve MU’s pricing leverage. Do not initiate solely on this report; trigger on confirmed construction or power-connection slippage.
• Set alerts for Korean policy decisions on industrial-power tariffs and state support. A formal tariff pass-through or sovereign-backed funding vehicle would remove the KEP short catalyst and could turn the current uncertainty into a positive for both KEP and Korean semiconductor capacity.