In the latter half of 2025, the semiconductor industry is poised to enter a new era. TSMC, the undisputed champion, and Samsung Electronics, hot on its heels, have ignited a fierce competition to begin mass production of the next-generation "2-nanometer (nm) process" that will serve as the core of future computing. As the explosive advancement of AI (artificial intelligence) reshapes the world's industrial structure, this is a decisive battle—one carrying geopolitical weight—for supremacy in manufacturing the most advanced chips, which will serve as the "brains" of that transformation. In determining the outcome of this battle, there is one core indicator that cannot be ignored: the "Yield Rate." No matter how excellent the design, it means nothing if it cannot be turned into products stably and at economical cost.

AD

The Dawn of the 2nm Era in Late 2025 – Three Distinct Strategies

The semiconductor industry's roadmap has set a major turning point for the latter half of 2025, as all the major players simultaneously declare their transition to next-generation processes.

  • TSMC (Taiwan): The unshakeable leader of the foundry (semiconductor contract manufacturing) industry. The company plans to begin mass production of its 2nm process in the second half of 2025. New fabs in Hsinchu's Baoshan and Kaohsiung will serve as the stage for this. Their biggest technical challenge is introducing, for the first time in this 2nm generation, a transistor structure called "GAA (Gate-All-Around)."
  • Samsung Electronics (South Korea): A comprehensive electronics manufacturer and perennial No. 2 in the foundry market. Samsung has set the exact same goal as TSMC—beginning 2nm mass production in the second half of 2025. The "Exynos 2600" application processor, rumored for its own smartphone "Galaxy S26" expected to launch in early 2026, is seen as likely to be the first product built on this node.
  • Intel (United States): The former emperor of semiconductors. Intel is betting its corporate fate on reviving its foundry business with a process that leapfrogs both TSMC and Samsung: "1.8nm" (which Intel calls "18A"). The company is also aiming for mass production in the second half of 2025, burning with determination to reclaim lost ground.

The fact that all three companies are converging on the same timeline for mass production of next-generation technology is highly unusual. This suggests that demand for chips aimed at AI and high-performance computing (HPC) has reached a point where it can no longer be delayed. And the mirror that most clearly reflects who is winning and losing in the early stages of this three-way battle is the yield rate.

The Watershed of Victory and Defeat: The Core Called "Yield Rate"

Yield rate refers to the proportion of manufactured chips that can be shipped as good products. This figure is the most important barometer for measuring a company's profitability, its ability to supply customers, and its technological maturity.

Putting together reports from multiple industry sources, the current situation for each company presents a stark contrast.

TSMC: A Stable "Over 60%"
According to industry reports, TSMC's 2nm process yield rate has already surpassed 60%, which is considered the benchmark for stabilized mass production. This is a remarkable figure. Despite introducing GAA architecture for the first time, achieving such a high initial yield demonstrates TSMC's overwhelming technological development capabilities and the depth of its manufacturing know-how. This stability is the primary reason why tech giants like Apple and NVIDIA place such enormous trust in TSMC.

Samsung: A Struggling "40%", Aiming for "Over 50%"
Samsung, on the other hand, finds itself in a more difficult position. Multiple Korean media outlets report that the company's 2nm process yield rate currently stands at around "40%." However, other reports indicate that after starting in the 30% range, the figure has continued to improve, and the company is now "aiming to achieve 50%." This reveals a picture of Samsung struggling yet desperately trying to overcome its challenges. Since mass production levels typically require a yield of at least around 70%, the road to reaching that target is far from smooth. This gap in yield rates vividly illustrates the current competitive disparity between the two companies.

This difference in yield rate carries more weight than mere numbers suggest. A lower yield rate means more wafers must be processed to produce the same number of good chips, causing manufacturing costs to soar. This not only puts a company at a disadvantage in price competition but also directly risks the inability to stably supply the quantities customers demand.

AD

The "GAA" Technology Architecture – Can Samsung Capitalize on Its First-Mover Advantage?

Another key protagonist in this technology race is the GAA (Gate-All-Around) architecture.

This new technology, which replaces the conventional FinFET structure, dramatically suppresses leakage current by surrounding the channel—the pathway for electric current—with the gate on all four sides, enabling higher performance with lower power consumption. As miniaturization approaches its limits, this is regarded as a trump card for achieving further performance gains.

What's interesting here is Samsung's strategy. The company introduced GAA ahead of its competitors, starting with its current 3nm process. This was an ambitious bet aimed at securing a "first-mover advantage." However, it is well known that they paid the price in the form of low initial yield rates.

For Samsung, this bitter experience with 3nm is meant to serve as a foundation for establishing manufacturing stability at 2nm. In a sense, the company is now in the phase of reaping the rewards of its early investment. If Samsung can fully master GAA know-how and dramatically improve its yield, there remains a possibility—however slim—that it could claim a technological advantage over TSMC.

TSMC, by contrast, prioritized stable supply at 3nm using "FinFlex," an improved version of the proven FinFET, before finally introducing GAA at 2nm when the time was right. This cautious, methodical approach—testing the waters before committing—has resulted in a high initial yield rate. It could be said that the difference in corporate culture between the two companies is reflected clearly in their strategies.

The Front Lines of the Customer Acquisition Race – Which Will Apple and NVIDIA Choose?

Ultimately, it is the customers who will render the verdict on this technology race. In particular, which foundry is chosen by tech giants leading the global industry—Apple, Qualcomm, NVIDIA, and AMD—will determine the balance of power within the industry.

According to market research firm TrendForce, TSMC commanded an overwhelming 67.6% share of the foundry market in the first quarter of 2025, while Samsung remained at just 7.7%. As these figures show, most major customers currently depend on TSMC.

TSMC Chairman C.C. Wei has stated that "demand for 2nm is ramping up at a pace exceeding that of the previous 3nm and 5nm generations," revealing strong interest from both the smartphone and HPC sectors. Market research firm Counterpoint Research also predicts that TSMC's 2nm production line will reach full capacity in the fourth quarter of this year, with a customer list that includes Apple, Qualcomm, MediaTek, AMD, and even rival Intel.

Samsung's biggest challenge is how to chip away at this stronghold. Unless the company can secure major external customers in addition to its own Exynos chips, recovering its massive investment will be difficult. The recent hiring of Margaret Han, a former TSMC executive, as head of Samsung's U.S. foundry business, likely reflects a serious strategic effort to break into major North American tech companies.

And with Intel joining this competition, the picture grows even more complex. Intel has announced that it secured major customers like Microsoft for its 18A process, and backed by national support, it is becoming a presence that neither TSMC nor Samsung can afford to ignore.

AD

Beyond the Miniaturization Race Lies "Supremacy in the AI Era"

This battle over 2nm is no longer simply about smartphones processing a bit faster. Its true essence lies in a struggle for supremacy over the future of industry itself—a contest over who will supply the computing power for the AI era.

  • Revisiting the Strategic Context:
    • TSMC: The strategy of a champion, wielding "trust and stability" as its weapon to maintain and expand a robust ecosystem with its customers.
    • Samsung: The strategy of a challenger, leveraging its position as the world's top memory maker alongside its foundry business to compete through comprehensive strength. Its head start in GAA technology may be its only path to a reversal of fortune.
    • Intel: The strategy of a veteran seeking revival through both design and manufacturing, riding the national tailwind of "reshoring manufacturing to the U.S."

The focus going forward will be on how much progress Samsung can make in overcoming its "Achilles' heel"—the yield rate—over the next six months to a year. If the company can establish a clear path to stable supply, customers may begin allocating orders to Samsung as a way to diversify risk and avoid excessive dependence on TSMC alone. Should that happen, market competition dynamics could kick in, potentially contributing to greater price and supply stability across the semiconductor industry as a whole.

What we are witnessing is not simply competition between companies. Against the backdrop of the massive paradigm shift that is AI, the future of the world's technology supply chain is being shaped by this battle unfolding in the infinitesimally small world of 2nm—nothing less than a historic moment. Whatever the outcome, one thing is certain: this fierce race in technological development will serve as the driving force accelerating our future.

Sources