The debate over who will lead the next generation of semiconductor manufacturing has reignited for a very specific reason: manufacturing yield performance. In the world of wafers, simply announcing a “smaller node” isn’t enough; what determines if a process is viable — and profitable — is the percentage of chips that come out functional from each wafer. And in this metric, market figures paint a more complex picture than headlines suggest.
In recent days, a report attributed to the investment firm KeyBanc has revived comparisons among the industry’s major players. According to this analysis, Samsung’s yield for its 2 nm GAA is still below 40%, a figure that, if true, would require the company to improve rapidly if it wants to turn its stake into a sustainable business.
The tension isn’t minor because Samsung has been through a delicate phase with its 3 nm GAA: the market still remembers the challenges related to industrial maturity and how those setbacks drove part of the fabless ecosystem to prioritize TSMC. Therefore, any data suggesting fragility at 2 nm is seen as a warning sign… though it’s important to highlight a key nuance: at these nodes, yield is a snapshot that changes month to month and depends heavily on the type of chip, its size, and the level of design optimization.
Why yield matters more than the “nanometer number”
When discussions mention a “yield below 40%” or “around 60%,” what’s really being debated is industrial economy:
- Cost per functional chip: fewer good units per wafer mean each chip costs more.
- Effective capacity: a factory can produce many wafers, but if the yield is low, useful production plummets.
- Commercial timelines: a node may be technically “real,” but commercially unviable if it doesn’t meet certain performance thresholds.
This point is crucial because the industry is entering a stage where leadership isn’t decided solely by lithography: it’s decided by consistency, capacity, and margin.
Samsung: doubts in a phase where other reports spoke of improvement
Although the conservative estimate cited by Wccftech places Samsung’s 2 nm yield below 40%, other industry reports in recent months have indicated a higher range, from 50% to 60%, and even an upward trend compared to the previous stage.
This divergence in figures illustrates a common pattern in semiconductors: yield estimates are rarely uniform, because they may measure different things (pilot batches versus ramp-up, small chips versus large ones, “by layer” yield versus final yield). Nonetheless, the strategic takeaway is key: Samsung needs to demonstrate to the market that its 2 nm isn’t just a technological improvement but an industrial platform capable of attracting volume and maintaining margins.
Intel 18A: the narrative of a comeback also depends on performance
Meanwhile, Intel is fostering its own comeback story with 18A, which combines new transistor architecture and significant changes in how the chip is electrically powered. In this context, there have also been reports placing Intel’s yield at around 60% for 18A. If confirmed, this would reinforce the thesis that the company can again compete at the forefront of advanced process technology.
Apart from specific percentages, market signals continue to attract attention: Intel has emphasized sustained improvements in its industrial capacity and ambitions to turn its foundry division into a strategic pillar. And in an environment where geopolitics weigh as much as technology, any credible progress in advanced processes within the U.S. is also seen as a factor of supply chain resilience, not just competitive advantage.
TSMC remains the benchmark, but the race is now about capacity and reliability
In this three-way comparison, TSMC is often seen as a reference point for industrial maturity and scaling. Some market analyses have reported higher figures (for example, around 80% in certain contexts), although, again, the exact number depends on the scope of the estimate.
What’s more tangible is that TSMC is already moving forward in 2 nm: the company has confirmed the start of production for its N2 technology at facilities like Baoshan, reinforcing the idea that the industrial timeline is progressing.
What clients and analysts should monitor in 2026
Rather than focusing on an isolated number (40%, 50%, or 60%), for a buyer — from mobile device manufacturers to AI integrators — the key factors in 2026 will often be:
- Yield trend (quarter-over-quarter), not the single data point.
- Supply risk and reserved capacity, especially if the market enters cycles of scarcity.
- Design ecosystem and support (EDA/IP), because on these nodes, every detail influences the final performance.
- Commercial signals: who commits to volume, which types of chips are prioritized, and delivery windows.
In other words, the “nanometer number” war is increasingly becoming a war of industrial execution.
Frequently Asked Questions
What does it exactly mean for a node to have a yield of 40% or 60%?
It indicates the percentage of functional chips obtained per wafer after completing the process. The higher this percentage, the better the cost per chip and the more viable large-scale production becomes.
Why are there such different estimates about Samsung’s 2 nm yield?
Because not all sources measure the same aspects: it can vary depending on the chip, die size, pilot phase or mass production, and the method used to estimate performance.
Can Intel 18A become a real alternative to TSMC and Samsung?
It can, if it sustains yields and capacity over time and proves reliability in production. Some market assessments point to significant improvements, but the decisive factor will be the industrial execution and commercial traction.
What should companies demand when choosing a foundry for advanced processes?
Beyond the node itself, they should prioritize secured capacity, a credible roadmap, design support, stable performance, and contracts with realistic delivery windows.
via: biz.chosun