The Lee Jae-myung administration is pushing to create a semiconductor cluster in Honam (Gwangju–South Jeolla) for Samsung Electronics and SK hynix under the policy line of balanced regional development. There has been no official announcement yet, but both companies are said to be reviewing plans to build new front-end process lines in Gwangju–South Jeolla, separate from the existing Yongin cluster. Some predict the investment could reach 1,000 trillion won.
The issue is that the actors steering a crucial decision that could determine the fate of corporations are the government and political circles, while the side that actually shoulders the expense is the corporations. Semiconductor equipment investment for a single production line requires tens of trillions of won and, once made, functions as fixed costs for decades, yet discussions on the Honam cluster are proceeding to fit the government's regional balanced development timetable, not on demand and technology trend outlooks for the global semiconductor market. Inside Samsung Electronics and SK hynix, there are concerns this decision could come back as an "expense bomb."
◇ Finance and fixed-cost risks… burden of running two clusters at once
The first concern is the medium- to long-term financial burden that will fall on Samsung Electronics and SK hynix. SK hynix has estimated it will cost 600 trillion won to build four phases of the Yongin cluster by 2050. On top of that, with Samsung Electronics already proceeding with up to 360 trillion won in investments, adding a Honam front-end investment could quickly stack up borrowing in a short period. If the cycle turns down, this burden risks being passed straight onto the corporations' balance sheet.
The financial burden leads to a complex calculus on fixed costs and utilization. If a front-end fab is added in Honam before the Yongin cluster even ramps up, depreciation and labor costs at the two bases will accumulate simultaneously. Semiconductors are an industry where fixed costs overwhelmingly outweigh variable costs, so when the cycle weakens, the bill for these fixed costs can reach corporations first.
Even a simple calculation shows the construction cost of a single memory fab is anything but small. The industry sees the recent cost to build one semiconductor fab at a minimum of 60 trillion won, and there are views that if both Samsung Electronics and SK hynix build front-end facilities in Honam, total investment could rise to 500 trillion won. By that measure, even if only one fab is built in Honam, new annual depreciation of 6 trillion won (based on 10-year depreciation) would arise regardless of whether it runs.
If a fab of that scale is added in Honam while tens of trillions of won are already being spent in Yongin and Cheongju, those fixed costs stack up as they are. Because these expenses do not fall even if factory utilization is lowered, the paradox in which unit costs soar as production is cut when the cycle turns down applies equally to a Honam fab. Moreover, this is a simple assumption reflecting only minimal expenses; in reality, depreciation periods and methods differ by company, and adding equipment aging and replacement expenses could make the burden heavier.
◇ Excess capacity, weaker technology, and downturn risks all around
Expanded capacity can return as a bearish factor in the market. In fact, past semiconductor "chicken games" were cases where overexpansion led to oversupply and a "battle royale" among memory companies. If the memory cycle is down when the Yongin and Honam clusters run simultaneously, the capacity both companies added could cannibalize their own price competitiveness. Price declines from excess capacity will become visible through impairments and inventory valuation in accounting.
From an efficiency standpoint, doubts remain that this is still a decision that defies understanding. Even if Gwangju–South Jeolla is assessed as capable of building an ecosystem for back-end (OSAT), it has virtually no base of materials, parts, and equipment partners needed for front-end processes. At a site newly designated by the government, initial production yield and productivity are likely to be lower than at existing capital-area fabs, which could show up in profit and loss as inventory and cost inefficiencies.
The long-term and less visible expense is delayed technology transition. Many of the research and development and production technology personnel at Samsung Electronics and SK hynix are based in the capital area, making it hard to staff a new base. Even the Yongin cluster is seeing delays in building the power grid due to some local governments' pending consent on substation installations; pushing Honam simultaneously could disperse management's decision-making capacity. That could be billed as an opportunity cost of missing the timing for next-generation process investments.
Concerns are already being raised inside and outside Samsung. Lee Chan-hee, Chairperson of the Samsung Compliance Committee, said, "We will closely watch to ensure semiconductor investment is not swayed by political logic without considering the impact on the sustainability of corporations and the national economy," noting it should not be subject to political logic. A senior Samsung official assessed, "In a worst-case scenario, if the peak of the memory supercycle overlaps with the start of operations at the new cluster, profitability could fall sharply."
Even if the push for a Honam cluster is formalized, it will take a long horizon of 10 to 20 years to reach actual operation. The expense of the decision the government makes now may ultimately be borne by the next management and by corporations in a downcycle. How the final investment size and schedule are coordinated between the government's push for speed and the corporations' caution will be the key point to watch.