Samsung Electronics' 2-nanometer foundry process has reached a yield rate above 60%, South Korean outlet Businesskorea reported in March 2026, citing industry sources, a level that has helped the company pursue customers including IBM and Baidu alongside its own Exynos chips.
Digitimes reported in August that Samsung's contract chipmaking business could approach full utilization in the second half of 2026, describing it as a potential turning point after years of underuse weighed on the division's earnings.
Why it matters
Samsung Foundry is the only large-scale alternative to TSMC currently shipping advanced logic in volume, alongside Intel's earlier-stage 18A ramp. A credible yield recovery would matter for customers who want geographic and vendor diversification away from TSMC, and for Samsung's own memory-adjacent AI chip ambitions, including HBM.
Tom's Hardware's August coverage tied the yield story directly to a $16.5 billion custom-chip deal with Tesla, illustrating how yield credibility translates into contract wins that Samsung badly needs to justify its Taylor, Texas and Pyeongtaek fab investments.
How it works
Samsung's 2nm effort uses gate-all-around transistor architecture, which the company adopted earlier than TSMC at the 3nm generation but has struggled to bring to acceptable yields until this round of improvements. Yield, the percentage of functional chips per wafer, is the single biggest driver of whether a foundry process is commercially viable, since low yields make each usable chip disproportionately expensive.
Samsung has also linked its foundry recovery to its HBM business: the Seoul Economic Daily reported that Samsung emphasized production stability while shipping HBM4, the newest generation of high-bandwidth memory, to Nvidia this year.
The competing read
Not all the recent reporting is optimistic. Tom's Hardware's August analysis described Samsung's foundry as still roughly a tenth the scale of TSMC's business and still fighting yield issues in absolute terms even after the improvement. And TrendForce reported on September 11 that a long-rumored deal to bring Qualcomm back to Samsung's 2nm line may slip to 2027, with sources citing Samsung's reluctance to lower pricing as its process stabilizes, rather than a pure technical failure.
That combination suggests Samsung's yield gains are real but not yet enough to win price-sensitive customers away from TSMC on favorable terms, a more mixed picture than the yield number alone would suggest.
What happens next
Samsung's ability to convert improved yields into signed volume contracts through the rest of 2026 will be the clearest signal of whether its foundry turnaround is durable. The Qualcomm decision, expected to become clearer over the next several months, is one of the most closely watched data points.
Samsung's Taylor, Texas fab and its HBM4 supply relationship with Nvidia are also likely to generate further disclosures as the company reports subsequent quarterly earnings.