Samsung's 2nm chip and the Korean chip renaissance

Samsung announced progress on their 2nm Gate-All-Around process, and the details suggest they’re closer to TSMC than most people assumed.

The race between these two companies is the most consequential industrial competition happening right now. More important than any AI company rivalry. More important than any social media platform war. Because whoever makes the best chips at the smallest scale controls the hardware layer that everything else runs on.

The state of the race

TSMC is ahead. That’s not controversial. They’ve been ahead for several years, and their N2 process is on track for volume production. Their customer list reads like a who’s who of technology: Apple, NVIDIA, AMD, Qualcomm, Broadcom. The world’s most important chips are TSMC chips.

Samsung has been the persistent second place. Their 3nm GAA process (the first to use Gate-All-Around transistor architecture) launched before TSMC’s equivalent, but yield issues kept them from capturing significant market share. The technology was right. The execution was lagging.

At 2nm, Samsung appears to be closing the gap. Their newsroom publications and industry presentations show power efficiency and performance metrics that are competitive with TSMC’s published targets. Whether those numbers hold in volume production is the real question, and it’s a question that won’t be answered until fabrication begins at scale.

Why Korea matters

South Korea has a chip manufacturing heritage that’s decades deep. Samsung and SK Hynix together produce the majority of the world’s memory chips (DRAM and NAND). Samsung’s foundry business extends this into logic chips, the processors and AI accelerators that run computations.

The Korean government is investing heavily. Tax incentives, infrastructure spending, talent programs. They understand that chip manufacturing isn’t just an industry. It’s a strategic capability. The same way oil-producing nations treat petroleum. The same way agricultural nations treat food supply.

Korea’s bet is that Samsung can compete with TSMC at the frontier, giving the world a second source for the most advanced chips. If Samsung succeeds, the concentration risk that keeps me up at night (90% of advanced chips from one company in one country) gets meaningfully reduced.

The technical challenge

2nm GAA is hard for everyone. Gate-All-Around means wrapping the transistor gate completely around the channel (a stack of silicon nanosheets). This gives better electrical control than FinFET (where the gate wraps around three sides) but introduces new manufacturing challenges.

The nanosheets need to be incredibly uniform in thickness. Variations of even a few atoms affect performance. The deposition, etching, and patterning steps require ASML’s most advanced EUV tools and process control at the atomic level.

Samsung’s historical challenge has been yield: the percentage of chips on a wafer that work correctly. At these scales, a single defect on a wafer can kill a chip. TSMC’s yield engineering is legendary, refined over decades of being the world’s preferred foundry. Samsung’s yields have traditionally been lower, which means higher effective cost per working chip.

If Samsung has solved (or significantly improved) their yield problem at 2nm, the competitive picture changes dramatically.

What I’m watching

I’m watching for customer announcements. TSMC’s dominance is reinforced by its customer relationships. If Samsung can win a major design win at 2nm (say, a Qualcomm chip or an NVIDIA product), that’s a signal that the gap has closed enough for demanding customers to trust Samsung with their most important designs.

I’m also watching the geopolitical dimension. In a world where Taiwan’s status is uncertain, having a competitive Korean foundry provides a strategic hedge. The US, Japan, and Europe are all investing in domestic chip manufacturing, but those fabs are years away from frontier capability. Samsung in Korea is a production-ready alternative that exists today.

The Korean chip renaissance isn’t just about Samsung vs TSMC. It’s about whether the world can have more than one source for the chips that power AI, phones, cars, and everything else.

I’m rooting for competition. Not because I want TSMC to lose. Because I want the world to win.

Related thinking: