Why SK's Chairman Warns of a Multi-Year Chip Shortage

The Signal Behind the Words

Choi Tae-won does not often step onto the international stage to forecast commodity crises. But in an interview with Japan’s Asahi Shimbun on September 8, the chairman of SK Group did exactly that—and the implications go well beyond Korea’s borders.

He warned that the semiconductor shortage driving the AI boom is unlikely to ease in the near term. He did not say it lightly: he cited NVIDIA CEO Jensen Huang confirming that GPU and system supply remains constrained by chip scarcity. Choi’s point was structural, not cyclical. Building a fab takes years. There is no shortcut to increasing output overnight, and none is visible on the horizon.

Then he dropped the real news. SK Hynix has ambitious investment plans in Korea, but they are not enough. The company is now evaluating locations worldwide for new semiconductor fabs—sites that offer land, power, talent, and an existing ecosystem. A decision on where to prioritize and how to proceed is expected soon.

That sentence alone shifts the map.

Why This Matters Now

Semiconductor capacity expansion has been a patchwork of national policies and corporate strategies. The U.S. CHIPS Act subsidizes domestic fabs. Europe is building its own. China is doubling down on self-sufficiency. Samsung and TSMC have been juggling competing priorities across Arizona, Kumamoto, and Songdo.

But Choi’s announcement carries a different weight because it comes from a leader of SK Hynix—the world’s largest memory chipmaker by revenue and the supplier at the heart of the AI infrastructure buildout. High-bandwidth memory (HBM) and advanced DRAM are the bottlenecks no one is talking about openly enough. Every generative AI model trains on them. Every data center expansion depends on them. And no one is producing enough of them.

The shortage is not just about volume. It is about a narrow set of products—HBM4, the next iteration of high-bandwidth memory—and a handful of suppliers. SK Hynix and Samsung are the dominant players. TSMC handles logic. That concentration creates fragility.

Choi is effectively saying SK will not wait for the problem to solve itself through Korean investment alone. It is looking outward, which means it is looking at options that range from the familiar (the United States, Europe) to the less discussed (Southeast Asia, the Middle East).

Japan Was Not an Accident

The most consequential part of the interview was not the shortage forecast or the fab search. It was Choi’s repeated emphasis on Korea-Japan cooperation in semiconductors.

He drew a sharp line between competition and complementarity. Korea accounts for more than 50 percent of global memory semiconductor sales. Japan’s share of finished products sits at roughly 5 percent. But flip the lens, and the picture inverts: Japan dominates materials, components, and equipment—the inputs that make fabrication possible. Tokyo Electron, Shin-Etsu Chemical, and Sumitomo Metal Mining are not配角. They are the spine of the supply chain.

This is the detail English-language coverage consistently flattens. The narrative around Korea-Japan semiconductor relations tends to reduce everything to geopolitical tension. But the industrial reality is far more interdependent. Japan supplies the photoresists, the specialized gases, the deposition tools. Korea turns them into the chips the world needs. Neither can do it alone at scale.

Choi named specific partners: NTT and Tokyo Electron. He said SK is actively exploring investment frameworks with Japanese counterparts, particularly around AI-related initiatives. That language—“exploring investment frameworks”—suggests structure beyond supplier contracts. It hints at joint ventures, co-investment, or shared R&D.

Why Japan matters for SK right now is straightforward. If Choi is serious about a multi-year shortage, then securing access to Japanese equipment and materials is as critical as finding land for a new fab. A fab without tooling is an empty building. A fab with tools but no domestic political support is a regulatory risk. Japan sits in the middle of both worlds.

Who Wins and Who Loses

SK Hynix wins by locking in supply chain advantage before competitors can. A global fab strategy gives it options—geographic and political—that Samsung currently does not have as clearly laid out. Samsung has committed heavily to Texas. SK is still free to diversify.

Japan wins if SK’s partnership moves beyond words. Japanese semiconductor equipment makers have been quietly rebuilding momentum since the early 2020s. More SK investment means more orders, more revenue, more leverage in a market where Taiwan and South Korea have traditionally held the scale.

The United States could lose if SK chooses locations outside American jurisdiction. CHIPS Act subsidies are generous, but they come with strings. A fab in Europe or elsewhere operates under different rules, different trade relationships, and potentially different export controls. That is a risk Washington is already watching.

Taiwan loses relevance only if SK diverts meaningful capacity away from its existing partnerships. So far, SK has not signaled a reduction in its Taiwan exposure. But the Fab 2 expansion in Hsinchu and new HBM lines there remain central to its strategy. This announcement is additive, not substitutive.

What Comes Next

The timeline Choi hinted at—“soon” for a decision—could mean announcements before the year ends. The market will be watching for three things: location, partners, and product focus.

A U.S. site would align with current subsidy structures but increase political risk given export control debates. A European site—likely in Germany or France—offers energy stability and EU support but faces a thinner talent pool. A Middle Eastern or Southeast Asian option would be the surprise play, trading ecosystem maturity for speed and cost.

The product mix matters more than geography. SK Hynix is racing to scale HBM production. Every new fab should be evaluated through that lens. If the priority is HBM, the site needs to be close to packaging lines and major AI customers. If the priority is DRAM volume, it needs cheap power and water.

Choi’s interview was carefully calibrated: a public signal to investors and governments, a private invitation to Japanese partners, and a message to competitors that SK is moving faster than expected. The semiconductor industry runs on those kinds of signals. The question now is whether the company behind them has the execution to match the rhetoric.

The shortage will not solve itself. The fab search is just the beginning.