1,600 Trillion Won Bet: Korea Told Fabs Alone Will Not Win

South Korea has 1,600 trillion won of private investment pledged across three mega projects — semiconductors, AI data centers and physical AI — yet experts warn the money will not deliver leadership without a broader industrial strategy.

The Presidential Advisory Council on National Economic Affairs held a public forum, "Three Mega Project: Tasks for Becoming a National Growth Platform," at Ferum Tower in Jung-gu, Seoul, on the 2nd. Vice Chair Kim Song-sik framed the stakes directly.

"The three mega projects are the biggest challenge since the 1970s heavy and chemical industrialization drive, fundamentally reshaping our economy and industries," Kim said. "Strategies should extend beyond expanding semiconductor fabs to strengthening the entire semiconductor ecosystem."

What did the government promise?

Industry Minister Kim Jung-kwan said the AI revolution has placed all nations on a new starting line. He called it a decisive opportunity for South Korea to become a global leader in future industries.

"The government's role is to remove obstacles for swift corporate investment and support essential infrastructure and systems," the minister said.

His plan links the three projects into one platform. Semiconductors supply AI computing power. AI data centers support AI learning and application. Physical AI applies the results to robots and factory floors to lift productivity. Gains would be reinvested into advanced sectors.

The infrastructure gap is measurable. As of July, 53 of 75 planned 345-kilovolt transmission line projects — all due for completion by 2030 — had not started. That is 70.7 percent of the pipeline. The government also wants to cut at least two years from semiconductor factory timelines through streamlined permitting and the Ultra-Wide Special Zone Act, which would grant R&D personnel exemptions from the 52-hour workweek.

What did industry demand?

Park Geon-su, vice president of the Korea Semiconductor Industry Association, said the market is a "winner-takes-all" structure. Technology leadership, timely mass production and customer acquisition reinforce the dominance of leading firms.

"Success depends not on announced investment totals but on whether investments start, fabs operate, and domestic supply chains are adopted as planned," Park said.

He called for:

• Preemptive power and water supply
• Support for domestic materials, components and equipment firms
• Swift enactment of the Ultra-Wide Special Zone Act with R&D work-hour exceptions
• Regional hubs: memory in the Southwest, advanced packaging in Chungcheong, materials and equipment in Daegu-Gyeongbuk and the Southeast, linked by a cross-ministerial management system

Baek Seo-in, an associate professor at Hanyang University, said South Korea must secure unique competitiveness because Taiwan and Japan are also nurturing semiconductor ecosystems. She proposed protecting small but strong domestic suppliers critical to economic security, promoting "science reshoring" to attract advanced R&D amid U.S. manufacturing reshoring, and jointly setting technology safety standards with major manufacturing nations.

Why is the memory lead not enough?

Lee Joon, head of the Strategic Industry Research Center at the Korea Institute for Industrial Economics and Trade, gave the forum its sharpest numbers. South Korea holds nearly 80 percent of the global high-bandwidth memory market. But non-memory sectors account for 76 percent of the global semiconductor market, where Korea lags.

Of the private money committed, 881 trillion won is concentrated in memory. Lee proposed combining memory and logic semiconductors to build system semiconductor competitiveness. He also called for a "Korean-style semiconductor innovation hub" where companies and universities collaborate on next-generation chips, modeled on Belgium's IMEC.

Can physical AI and data centers deliver?

Park Min-won, president of Changwon National University, said robots must learn and use skilled manufacturing techniques. If overseas tech giants train AI models on Korean manufacturing data and skills, the country could lose technological leadership. He proposed step-by-step training: robots learn basic movements and process-specific tasks, prove out in pilot factories, then deploy to unmanned plants. He also urged clear data usage rights and compensation for skilled workers.

Kong Kyung-chul, a professor at KAIST, pushed robot applications beyond humanoids into agriculture, disaster response, healthcare and caregiving. He called for national investment in actuators, control systems and force and tactile data collected by robots.

For AI data centers, the demand gap is the core problem. Jung Byung-sun, a professor at Sogang University, cited the 2035 target of 18.4 gigawatts of AI data center capacity. Confirmed demand under long-term contracts stands at only 2.5 GW. He recommended prioritizing the first-phase 8.4 GW by 2029 for projects with guaranteed demand, then approving the remaining 10 GW sequentially based on customer contracts and grid connectivity.

Yoon Dae-gyun, director of the AI Research Institute at Ajou University, warned against passive ownership. "If AI data centers are merely leased to foreign companies, operational expertise will accumulate overseas," he said. He pressed for domestic AI models and semiconductors inside those facilities.

The forum's verdict is likely to shape policy through the decade: with 70 percent of grid projects unstarted and barely an eighth of targeted data center demand contracted, execution — not headline investment figures — will decide whether Korea converts 1,600 trillion won into durable industrial leadership.