SK's Gamble: Racing China in Semiconductors While Deepening AI Ties

A Warning From Shanghai

Choi Tae-won did not mince words. Speaking to reporters at a business summit in Shanghai on October 11, the chairman of SK Group delivered a stark assessment of the global memory chip landscape: “If China or American company Micron expands capacity faster, our market share will shrink.” His prescription was blunt — SK must invest aggressively now or risk falling behind irreversibly.

The warning carries unusual weight for a South Korean conglomerate leader. Choi’s public urgency marks a departure from the traditionally cautious capital allocation that has defined Korean industrial strategy. What he is describing is not merely a competitive challenge — it is a structural inflection point in the global semiconductor order.

The Chinese Threat Is Real and Accelerating

Choi’s concern is anchored in hard data. China’s CXMT and YMTC are ramping DRAM and NAND flash production capacity at a pace that would have been unthinkable five years ago. Backed by massive state subsidies and access to the world’s largest semiconductor end-market, these companies are closing the technology gap while Korean firms face their own constraints.

The irony Choi highlighted is sharp: “We cannot expand in the Chinese market, making it difficult to compete properly.” SK and its peers are effectively locked out of scaling production in China due to export controls and geopolitical friction, even as Chinese rivals feed off domestic demand to achieve economies of scale.

This dynamic creates a structural disadvantage that no amount of efficiency improvements can fully offset. Memory chip markets are notoriously cyclical and volume-driven. Whoever commands the largest capacity at peak demand cycles sets the price — and China is building capacity regardless of cycle conditions, subsidized by state policy.

The 400 Trillion Won Response

SK’s answer is a dramatic acceleration of domestic expansion. The company revealed in June a plan to build a 400 trillion won semiconductor cluster in Gwangju, roughly two-thirds the scale of its existing Yongin facility. Choi confirmed on-site that the Yongin construction schedule has already been compressed by twelve years and that parallel construction at both sites may be necessary.

He signaled willingness to break ground before supporting infrastructure is fully complete — a significant departure from the usual practice of waiting for power and water facilities to be ready. The message is that SK views the timeline as unforgiving.

Japan was also flagged as a potential site for overseas expansion, though details remain scarce. The consideration of multiple geographic bases reflects a company betting that the next generation of memory chip demand will be so large that no single site can serve it alone.

The AI Counterbalance

Here is where the strategy gets contradictory — and interesting. While Choi warns about Chinese competition in memory chips, SK is simultaneously expanding its AI research presence in Shanghai. The company plans to establish an AI R&D center there focused on data centers, physical AI, and biopharma applications.

This dual posture — competing aggressively in semiconductors while deepening technology cooperation in AI — may seem inconsistent. But Choi framed it as coherent: semiconductors are a speed race, he said, while AI is a cooperation game.

The logic, if it holds, is that SK can compete on chip manufacturing scale while sharing AI frontier research with Chinese partners. It mirrors the broader pattern of Korean and Japanese companies navigating US-China decoupling pressures — maintaining commercial ties where they can while competing where they must.

Choi pushed back explicitly against US political rhetoric questioning whether Samsung and SK should operate in China. “I do not think it is very wise to believe that restricting Korean companies’ production activities in China would prevent China’s semiconductor industry from growing,” he said. The argument is pragmatic: cutting ties will not slow Chinese advancement, but it will cede ground.

The Social Value Expansion

There is a third strand to Choi’s strategy that deserves attention. Over the past decade, he has built out SK’s social value framework, including the Social Performance Contract (SPC) system that rewards enterprises for measurable social impact. The Social Value Research Institute is now taking this model international, hosting a forum with fifteen countries participating.

Choi argued that traditional corporate social responsibility falls short because it focuses on spending and publicity rather than outcomes. AI, he suggested, can finally make precise measurement of social impact economically viable. His vision is a “New Capitalism” where profit-seeking and social value creation are both rewarded — a framing designed to attract capital from investors who increasingly demand ESG alignment alongside financial returns.

Whether this amounts to genuine transformation or sophisticated positioning remains to be seen. But the timing — coinciding with SK’s most aggressive industrial bet — suggests Choi is preparing the company for a future where capital allocation decisions face scrutiny on both financial and social metrics simultaneously.

What This Means for the Global Chip Market

SK’s announcements carry implications that extend far beyond Korea. If Choi’s read is correct — that the window for maintaining memory chip leadership is narrowing — then other Korean and Japanese players may face similar pressure to accelerate. A wave of new fab announcements could follow, potentially reshaping capacity dynamics in an industry that has spent the past decade deliberately limiting excess.

The Gwangju cluster, if realized at projected scale, would be one of the largest semiconductor investments in Korean history. It would also deepen the concentration of SK’s memory operations in a single region, creating both efficiencies and risks.

For the United States, Choi’s comments represent an implicit challenge to the export control strategy. He is arguing that restricting Korean companies’ Chinese operations will not stop Chinese chip advancement but will instead weaken the competitive position of America’s closest allies in the semiconductor supply chain. That is a message Washington may need to hear as the debate over technology decoupling intensifies.

The memory chip market is entering a period of extraordinary demand driven by AI training workloads. SK’s calculation is that the winners will be those who can bring capacity online fastest — not those who can operate most efficiently within constraints. It is a high-stakes bet, and the first mover advantage in memory is both real and temporary. Whatever SK builds today will define its position for a decade.

Choi’s Shanghai speech was not just a corporate strategy announcement. It was a signal that the oldest assumption of the Korean semiconductor industry — that caution and precision would win — may no longer be sufficient. The speed race has begun.