Samsung Electronics and SK hynix have rejected a proposal from Korea Electric Power Corp. to prepay about 25 trillion won, or roughly $19 billion, in electricity bills, according to a government document reviewed by Reuters. The decision turns what looked like an inventive financing plan for South Korea’s semiconductor expansion into a harder question: who should fund the grid before the country’s largest chip factories begin consuming the power?
KEPCO proposed the arrangement this month as an alternative to issuing more debt. Korean reports said Samsung was asked to advance 20 trillion won and SK hynix 5 trillion won, amounts based on approximately five years of their recent electricity bills. The utility would use the cash to accelerate transmission and substation construction for semiconductor clusters, then offset future power charges against the prepaid balances while compensating the companies with interest.
Reuters reported Monday that the chipmakers concluded after internal reviews that such large upfront payments would be difficult. A government official cited uncertainty about the durability of long-term semiconductor demand. That reasoning deserves care. It does not establish that Samsung or SK hynix expects the artificial-intelligence investment cycle to collapse. It does show that even cash-generative chipmakers are unwilling to treat a five-year demand forecast as certain enough to lock billions of dollars into a utility-financing structure.
The distinction matters because the proposed prepayment was never merely an electricity bill. Economically, it would have made two industrial customers important financiers of the state-controlled grid operator. KEPCO said when the proposal emerged that participation, amounts, rates and duration had not been finalized. The structure promised benefits to both sides: KEPCO could reduce bond issuance, while the chipmakers could improve the odds that power infrastructure would arrive when new fabrication plants need it. The rejection suggests the price, liquidity cost or allocation of risk was not compelling enough.
KEPCO’s financing constraint remains. Seoul Economic Daily reported that its consolidated debt reached 210.7 trillion won at the end of June, up from 205.6 trillion won at the end of 2025. The utility earned 4.91 trillion won in operating profit during the first half but incurred 2.1 trillion won in interest expense. It has planned 72.8 trillion won of grid investment through 2038, while a temporary expansion of its statutory bond-issuance ceiling is due to expire at the end of 2027.
Those figures explain why prepayment appealed to KEPCO. Pulling forward future electricity receipts could finance construction without immediately adding the same amount of conventional borrowing. It would not, however, create immediate revenue or profit because the payments would be recognized as electricity is delivered. It would chiefly reshape the timing and source of the utility’s cash.
For South Korea, the stakes extend beyond one rejected proposal. In June, Samsung and SK hynix said they would invest a combined 800 trillion won in a new chipmaking hub in the country’s southwest, with each company planning two fabrication plants, according to the Associated Press. Those commitments sit alongside the existing Yongin cluster and a national strategy built around semiconductors, AI data centers and related infrastructure. Factories can be announced faster than transmission lines can be permitted, financed and built.
The refusal therefore exposes a mismatch at the center of the AI capital cycle. Chipmakers want infrastructure delivered on schedule, but they also need flexibility if demand, technology or production economics change. KEPCO needs capital years before the new load produces ordinary electricity revenue. The government wants strategic capacity without allowing grid finance to become a bottleneck.
The next proposal will probably have to distribute that risk more explicitly. Government capital, revised electricity tariffs, project-specific financing, guarantees or a smaller customer contribution could all form part of a replacement. What Monday’s rejection makes clear is narrower but important: enthusiasm for long-term AI demand does not automatically translate into a willingness to finance five years of power consumption in advance. South Korea’s semiconductor ambition now depends as much on credible infrastructure finance as on the scale of its factory plans.