SK Hynix raises $26.5B in the biggest foreign IPO in US history, is urged to build new US fabs

Friday's $26.5 billion debut didn't just break records. It rewrote the rules for how Asian tech giants access American capital. SK Hynix priced its ADRs at a 2.7% premium to its Seoul average, then watched demand swamp the offering seven times over. The stock opened 14% above IPO price and kept climbing. This wasn't a polite market welcome. It was a feeding frenzy.

The Korea Discount — that persistent valuation haircut investors applied to Korean firms citing governance opacity, chaebol cross-holdings, and the North Korea overhang — just vaporized for the one company that matters most to the AI build-out. SK Hynix makes high-bandwidth memory. Nvidia's Blackwell and Rubin architectures choke without it. The market didn't price a Korean company. It priced a choke point.

Wall Street understands leverage when it sees it. The proceeds flow three ways: a new Korean fab already rising to address the global memory shortage, a packaging facility in the same country, and EUV scanners — the only machines that can print next-generation nodes. Notice what's missing: not a single dollar earmarked for American soil. The IPO prospectus makes that explicit.

Howard Lutnick noticed too. The Commerce Secretary showed up at a Micron event Thursday with a message that sounded less like invitation than instruction. He's "in talks" with both SK Hynix and Samsung about US fabs. Translation: the administration wants memory sovereignty on American territory, and it intends to use the CHIPS Act and tariff leverage to get it. Micron, reading the room perfectly, pledged $250 billion for US manufacturing and 90,000 jobs. That's not investment. That's political alignment.

The timing exposes the tension. SK Hynix and Samsung together committed over $550 billion for new Korean capacity — a figure that dwarfs Micron's US pledge. They're building where the engineering talent clusters, where the supply chains matured over three decades, where the government backs them with infrastructure and tax policy. Now Washington wants them to duplicate that ecosystem six thousand miles away.

Duplication costs more than money. It costs time — the one resource the AI arms race doesn't have. A greenfield fab takes four to five years from groundbreaking to qualified output. EUV tool lead times stretch past two years. The engineers who know how to tune HBM yields don't live in Idaho or Ohio. They live in Icheon and Giheung. Moving production means moving people, or training new ones from zero. Both bleed velocity.

Lutnick's pressure creates a genuine strategic dilemma for SK Hynix. Refuse the US ask, and you risk tariffs, CHIPS Act exclusion, and the "unreliable partner" label that killed Huawei's western access. Accept, and you fragment your own roadmap, dilute your best engineers across continents, and hand Samsung a home-field advantage in Korea. Samsung, notably, hasn't IPO'd in New York. It doesn't need to. Its balance sheet funds its own expansion. SK Hynix just sold a piece of itself to American investors who now expect returns — returns that depend on execution speed, not political theater.

The IPO proceeds buy EUV scanners. That's the tell. ASML's High-NA tools are the bottleneck for every sub-10nm node. SK Hynix is securing its allocation before the queue lengthens further. If Lutnick wants those scanners running in New York instead of Cheongju, he'll need to guarantee ASML delivery slots that don't exist. The Dutch don't take political IOUs.

Micron's $250 billion announcement serves a purpose beyond capacity. It sets a rhetorical floor. Every subsequent Korean commitment gets measured against it. But Micron's money spends over a decade. SK Hynix and Samsung's half-trillion-plus spends over a similar horizon. The numbers are theater. The only real metric is qualified wafers per week in 2027. On that metric, Korea leads. The fabs exist. The yields are proven. The HBM3E stacks are shipping today into Nvidia racks.

American investors who bought Friday aren't betting on US fabs. They're betting on HBM monopoly rents. They priced the ADRs knowing the Chips Act Subsidies may never materialize for foreign owners, knowing the Korea Discount could return if geopolitics flare, knowing the company's capital allocation committee answers to a board in Seoul. They bought anyway. That's not faith in American manufacturing policy. That's faith in physics — AI needs memory, and SK Hynix makes the best version money can buy.

The editorial lesson: capital follows choke points. Policy chases capital. The $26.5 billion isn't a vote of confidence in American industrial policy. It's a down payment on the only memory architecture that works for the next GPU generation. Lutnick can ask for fabs. He can't legislate yield curves. SK Hynix will build where the physics works. If that happens to be America, fine. If it stays Korea, the ADR holders still get paid. The market already decided which outcome it's pricing.