Success of three megaprojects hinges on execution

Chung Un-chan

The author, a former president of Seoul National University and former prime minister, is the chairman of the Korea Institute for Shared Growth.

The government recently unveiled three “megaprojects” centered on semiconductors, AI data centers and physical AI. Samsung and SK will build semiconductor fabs in Gwangju, while high bandwidth memory packaging capacity will expand in the Chungcheong region, including Cheonan and Onyang. Gumi, North Gyeongsang, will become a robotics hub, while large AI data centers are planned for Ulsan, Donghae and Sejong.

The vision is to link semiconductors, computing infrastructure, robotics and manufacturing into a single value chain, then combine it with the “five hubs, three special zones” strategy for balanced regional development.

The government plans to prioritize fiscal resources for the projects and establish a dedicated AI unit at the presidential office, with President Lee Jae Myung personally overseeing progress.

There is debate over the projects’ merits. Rather than address that debate, I would like to suggest how such an ambitious national strategy should be implemented.

First, the larger the blueprint and resources involved, the more meticulous execution must be. If national strategy becomes driven by political slogans or speed for its own sake, it risks overlooking the complexity of the real economy, producing inefficiency and eroding trust.

The announced private investment plans amount to several quadrillion won. One trillion won itself is enormous. Even if Jesus or Buddha had spent 1 million won ($708) every day from their birth until today, neither would have come close to spending 1 trillion won. Korea’s annual central government budget also is well below that figure.

Successful investment on this scale could transform Korea’s industrial and geographic structure. But faulty demand forecasts or excessive competition among regions could produce duplication and overinvestment.

Korea has been here before. Heavy and chemical industry development in the 1970s laid the foundation for today’s manufacturing competitiveness. Yet overlapping investment among conglomerates and poor demand forecasts, compounded by global oil shocks, left factories operating at only 30 to 50 percent of capacity by the decade’s end. Without the boom of the late 1980s, that excess capacity might have burdened the economy much longer.

In the 1990s, loose government-led financial policies encouraged companies to borrow short-term funds overseas and aggressively expand facilities. Corporate and financial-sector weaknesses accumulated, triggering cascading bankruptcies and helping set the stage for Korea’s 1997 bailout by the IMF. The lesson is clear: Large investments driven by optimistic expectations rather than market discipline can pose enormous risks.

Second, investments with high uncertainty and irreversibility require flexibility. Semiconductor fabs and data centers are expensive and difficult to repurpose once built. AI data centers are particularly sensitive to technological change because they depend on specialized assets such as substations, advanced cooling systems, high-speed networks and costly GPUs.

Rather than committing everything at once, policymakers should preserve the option to adjust the timing and scale of investment as information accumulates.

The megaprojects should therefore follow a phased, conditional investment model. Core infrastructure can be built first, while facilities expand according to demand, technological developments and changes in the power grid. Data centers should be modular, allowing expansion or conversion. Fiscal support, tax incentives and policy financing should also require companies to bear responsibility and demonstrate profitability.

Third, the roles of government and markets must be clearly defined. Rather than intervening in individual companies’ decisions about factory locations or scale, the government should concentrate on new social overhead capital, including transmission grids, water supplies, data-sharing systems and computing infrastructure.

Companies should retain autonomy to invest according to demand and profitability. At the same time, such infrastructure must not become the exclusive property of particular corporations. Korea needs an open ecosystem in which small- and medium-sized enterprises, startups and universities, especially regional universities, can access high-performance computing resources.

Finally, human infrastructure matters as much as physical infrastructure. A fab can be built within several years, but developing skilled workers and creating communities where they and their families will settle takes much longer.

The three megaprojects could become a foundation for Korea’s sustainable economic advancement. But success will depend on avoiding duplication and overinvestment, respecting market discipline and applying the philosophy of shared growth. The scale of ambition matters. But how Korea executes it will matter even more.