Samsung, which overtook TSMC at 10nm... Why did it fall behind? [Tech Talk]

Samjeon, World's First 10nm Mass Production in 2016
Gap Widens in Apple-TSMC Honeymoon Relationship
Starting from This Year, Technology Reset... Samjeon and Intel Strive
A Fierce Competition Where No Clear 1st Place Can Be Predicted May Unfold

Samsung Electronics Foundry was established in 2005. Since then, it has engaged in fierce competition with Taiwan's TSMC, the industry leader in semiconductor contract manufacturing, over market share and nanometer (㎚) technology.

Generally, TSMC is considered technologically superior to Samsung Electronics, but in fact, Samsung once surpassed TSMC in advanced process technology. Nevertheless, the market share gap between the two companies has only widened. Why did this happen?

Samsung Electronics, One Year Ahead of TSMC, Nullified by the 'Apple Effect'

Samsung Electronics Semiconductor Factory
[Image source=Samsung Electronics]

In 2016, Samsung Electronics operated the world's first 10㎚ foundry production line, ushering in the era of so-called 'leading-edge processes.' TSMC only began 10㎚ production in 2017, a year later than Samsung.

At that time, the semiconductor manufacturing industry faced technological barriers. Achieving 10㎚ required ASML's new EUV lithography equipment, but due to the complexity of the machinery, delivery schedules were delayed.

Samsung Electronics and TSMC have been competing technologically for dominance in nanometer (nm) fine process technology.

Samsung surpassed TSMC by pushing the limits of older DUV lithography equipment using 'triple patterning' technology. Previously, TSMC was at the 14~15㎚ process node, while Samsung was at the 20㎚ level, so Samsung achieved a 'quantum jump.'

However, the market share gap between the two companies did not narrow. In fact, the most significant event that reshaped the foundry market in the late 2010s was not Samsung's world-first 10㎚ process, but the birth of the Apple-TSMC alliance, which had a much greater impact.

Apple Partnership Completes TSMC's Unmatched Lead

Apple iPhone 15 [Image source=Apple]

In 2017, Apple entrusted the entire production of its self-designed semiconductor AP (Application Processor) 'A10' to TSMC's 10㎚ process. Previously, Apple had balanced orders between TSMC and Samsung, but from this point, a close relationship between Apple and TSMC was formed.

Apple is just one of many fabless semiconductor design companies, but its order volume is so massive that it shifts the market's balance entirely. As of last year, Apple alone accounted for $16.5 billion, or 23%, of TSMC's $72 billion revenue.

In the same year, this amount was close to 80% of Samsung Foundry's $20.8 billion revenue. Losing Apple alone causes the gap between TSMC and Samsung to widen significantly.

Transistor Density, Not Nanometers, Is Key

Transistor integration density is a key criterion for gauging semiconductor performance. [Image source=Yonhap News]

Why did Apple choose to partner with TSMC, which was a step behind Samsung at the time? There were likely various behind-the-scenes calculations. First, Apple probably did not want to share its core designs with Samsung, a competitor in the smartphone market.

But there was also a clear performance gap between Samsung and TSMC. In fact, the nanometer numbers that manufacturers publicize are not very meaningful when gauging the technological level of the latest semiconductors.

Even at the same nanometer scale, the 'transistor density'?the number of transistors per unit die area, which forms the basis of semiconductor performance?varies greatly depending on the manufacturer's expertise. For this reason, in 2019, Philip Wong, TSMC's Vice President of Research, pointed out that "(the nanometer number) is just like a car model number and should not be confused with actual content."

Since the late 2010s, Samsung has lagged somewhat behind TSMC in transistor density. For example, Samsung's 5㎚ 5LPE process integrates 126 million transistors per square millimeter of die, while TSMC's 5㎚ N5 process integrates 173 million, about a 37% difference.

A higher transistor density means faster semiconductor speed and improved power efficiency. Even if Samsung and TSMC manufacture chips from the same design, there will be a significant performance difference.

Technology Leadership Returns to Square One... Will the Foundry Battle Intensify?

Pat Gelsinger, Intel's Chief Executive Officer (CEO), unveiling Intel's 18A wafer. [Image source=Intel YouTube]

The semiconductor industry's characteristic is rapid changes in technology trends. This means there is no guarantee that the industry leader will remain the technological strongest forever. In fact, Samsung has once again succeeded in leading TSMC in the mass production competition of 3㎚ chips.

Samsung also achieved the feat of introducing the next-generation transistor GAA (Gate-All-Around) earlier than TSMC. TSMC plans to use the existing FinFET transistor up to the 3㎚ process and apply GAA technology starting from the 2㎚ process. For Samsung, now is the perfect time to attract new customers.

The rapid pace of technology trends means the two-way competition between TSMC and Samsung can change at any time. Intel recently made a surprise announcement at its 'Innovation 2023' event that it will produce chips using the 2㎚ (Intel's 20A) process next year. In 2025, it will introduce an even more advanced 18A (1.8㎚) process.

Currently, Intel's foundry service has reached the 'Intel 4' process (formerly 7㎚). Although this seems far behind Samsung and TSMC in terms of nanometers, its transistor density can already rival TSMC's 5㎚ process.

Therefore, if Intel realizes the roadmap it announced, it will be the first time that Samsung and TSMC's technological dominance is broken. Adding to this, including China's SMIC, which recently overcame U.S. technology sanctions to succeed in its own 7㎚ process, the contract manufacturing industry may enter a Warring States period where the competitive landscape is difficult to predict.

This content was produced with the assistance of AI translation services.