Korean Chipmakers Enter US R&D Circle Alongside Rivals

The Seat at the Table Has Changed

For decades, South Korea’s semiconductor strategy ran on a simple premise: build the best memory chips at the lowest cost, ship them fast, and let the fabs speak for themselves. Samsung and SK Hynix have dominated DRAM and NAND production through sheer scale and process discipline. But scale alone is no longer the differentiator.

In October 2026, AMAT revealed the EPIC center in Sunnyvale, California—a $5 billion (6.6 trillion won) joint R&D facility that changes the geometry of the entire industry. Samsung and SK Hynix are founding partners alongside TSMC, Intel, and Micron. They are not customers renting lab space. They are co-investors in the architecture of next-generation chip manufacturing.

This is a quiet but significant repositioning. Korea is moving from being the world’s most efficient fabricator to being a stakeholder in the R&D infrastructure that will define what gets fabricated—and where.

What EPIC Actually Is

The facility spans roughly 55,700 square meters, with about 16,200 square meters dedicated to cleanroom space for advanced process research. It was designed from the ground up to collapse the traditional separation between equipment development and chip manufacturing. In the old model, AMAT or Applied Materials would build a tool, hand it to a fab, and the fab would spend months or years optimizing it. The EPIC center is meant to run those processes in parallel—material science, equipment design, process development, and pilot production all under one roof.

The rationale is straightforward enough. AI-driven compute demands have pushed process complexity to levels where sequential development cycles are no longer viable. As AMAT CEO Gary Dickerson put it, the pressure from high-performance computing applications has introduced unprecedented complexity across the entire manufacturing chain.

What makes the announcement notable is not the facility itself but who is inside it alongside Samsung and SK Hynix.

The Company List Reads Like a Geopolitical Map

TSMC’s presence is the most politically consequential element. Taiwan’s foundry leader, which produces the vast majority of the world’s most advanced logic chips, is now sharing R&D infrastructure with Korean memory makers and with Intel—the American giant trying to reclaim its manufacturing edge. They are not competing in this room. They are coordinating.

Micron’s participation adds another layer. An American memory maker sitting next to two Korean memory makers in a California facility funded partly through CHIPS Act-adjacent partnerships signals that Washington sees memory supply chain resilience as a strategic priority equal to logic.

For Samsung and SK Hynix, the calculus is different but equally strategic. Both companies face intensifying competition from Chinese memory producers and from TSMC’s aggressive expansion into memory-adjacent markets like HBM. Being inside EPIC gives them early access to equipment development cycles that they previously accessed only after AMAT released products to the market. For HBM production—an area where SK Hynix currently leads and Samsung is racing to catch up—that timing advantage is commercially meaningful.

What This Means for Seoul-Washington-Taipei Dynamics

The EPIC center formalizes a dependency that already exists but was never this explicit. Korean chipmakers depend on American equipment. American fabless companies depend on Korean memory. Taiwanese foundries depend on both. EPIC makes that triangular dependency into shared infrastructure.

From Washington’s perspective, this is exactly the kind of alliance architecture the CHIPS Act was designed to encourage—pulling key players into US soil, giving American institutions influence over the pace and direction of technology development, and reducing the risk that advanced process knowledge concentrates entirely in Taiwan or South Korea.

From Seoul’s perspective, there is a trade-off. Participating in a US-led R&D hub deepens technological alignment with Washington, which has been a consistent Korean policy goal since the CHIPS Act negotiations. But it also means that breakthroughs developed at EPIC will be shaped by the priorities of American partners as much as by Korean ones. Samsung and SK Hynix gain influence by being in the room, but they also cede some autonomy over the direction of process development.

For Taipei, the implications are more delicate. TSMC’s participation signals willingness to integrate into the US alliance framework, but it also normalizes the idea that Taiwan’s most advanced capabilities should be co-developed on American soil. That is a long-term strategic risk that Seoul is arguably more insulated from, given that Korea already hosts massive Samsung and SK Hynix fab operations.

Who Wins, Who Loses

Samsung and SK Hynix win access. The speed at which they can iterate on HBM and advanced DRAM processes improves when they are working side-by-side with equipment vendors during the development phase rather than reverse-engineering finished tools. Intel wins by pulling a leading memory competitor into its ecosystem. TSMC wins by keeping Korean memory developers close enough to coordinate but far enough to maintain foundry independence.

The losers are less visible but real. Chinese memory producers like YMTC and ChangXin get further excluded from the inner circle of equipment-process co-development. SMIC, already sanctioned from leading-edge US equipment, is now structurally further away from the feedback loop that turns equipment innovation into process innovation. The gap between those inside EPIC and those outside it will widen.

What Happens Next

The EPIC center will begin operational testing in phases. Stanford and UC Berkeley are listed as research partners, which means the pipeline from academic research to industrial application will run through this facility—another layer of American institutional capture that benefits the host country in ways that go beyond immediate chip output.

Samsung’s CTO Song Jae-hyuk described the center as a way to accelerate the full arc from innovation to production. SK Hynix president Kwak Noh-jung specifically referenced HBM, DRAM, and NAND as areas of focus. TSMC’s Morris Chang-era successor, C.C. Wei, highlighted advanced packaging. Micron’s Sanjay Mehrotra pointed to shortened development cycles for next-generation memory.

Each company is announcing the same language—acceleration, collaboration, speed to production—but the underlying motivations differ. For Korea, the play is about securing a voice in the infrastructure that will underpin the next decade of semiconductor advancement. For the US, it is about ensuring that voice is exercised on American soil.

The old model had Samsung and SK Hynix winning by building faster. The new model rewards those who help decide what gets built—and where it gets built. EPIC is the first major test of whether Seoul can play that game.