hankyoreh

South Korean President Yoon Suk-yeol (left) stands for a photo with US President Joe Biden (center) and Japanese Prime Minister Fumio Kishida (right) while at the APEC summit in San Francisco, US, on Nov. 16. (Yonhap)

Semiconductors are a key strategic industry in the era of “technopolitics,” a term that describes how geopolitical hegemony is determined by technology. Kwon Seok-joon, professor at Sungkyunkwan University and author of the book “Semiconductors of the Three Kingdoms,” analyzes the significance of the US and China’s competition for chip supremacy, with its focus on national security, and the importance of the year 2024, with its packed political schedule.

It’s important to recognize that the true purpose of the US government’s push to reorganize the global semiconductor industry goes beyond economic gain and lies in recalibrating values from the perspective of national security.

This past September, I met with US Assistant Secretary of State Ramin Toloui while he was visiting Seoul. In conversations with former high-ranking officials and think tank analysts at various forums, they’ve all said more or less the same things.

As is widely known, the long-term strategy of the US government goes beyond just checking China. Indeed, the mainstream view is that containing China is an inevitable byproduct of the US’ campaign to reorient the semiconductor sector on itself and increase the share of cutting-edge chipmaking that takes place in the US.

High-tech semiconductor production is heavily concentrated in the countries of Northeast Asia — including South Korea, Taiwan and China — to a degree that the US regards as excessive. The US is concerned this overdependence will only increase with time because of China’s growing share of semiconductor production.

One of the US’ major long-term strategies is reducing or dispersing the dependence on Taiwanese chipmaker Taiwan Semiconductor Manufacturing Company, or TSMC, which produces a large share of logic chips (non-memory semiconductors).

Nvidia, which currently relies on TSMC for the majority of its GPU (graphics processing unit) production, has indicated it’s considering a foundry other than TSMC and Samsung Electronics, its primary competitor. While Nvidia didn’t get into the specifics, it’s presumably talking about American chipmaker Intel.

The overdependence on TSMC, which controls at least 85% of the production of logic chips at the size of 10 nanometers (nm) and below, is a big problem.

That’s because TSMC’s state-of-the-art processes are responsible for producing most advanced logic semiconductors, including ones for the central processing units (CPUs) that go into computers; the latest dedicated application processor (AP) chips for smartphones; artificial intelligence accelerator GPUs, TPUs, and NPUs; and data processing units (DPUs). Indeed, the situation is a near-monopoly.

The decision by TSMC to abandon its unspoken policy of producing advanced semiconductors only within Taiwan and to commit a major investment in the US city of Phoenix — more than US$40 billion, or roughly triple its original plan — stems from fears about this sort of US backlash and the potential for a forcible “adjustment” to its monopoly.

Around the time of the groundbreaking ceremony for the Arizona plant, TSMC founder Morris Chang grumbled about free trade being “dead.” This situation offers some insight into why he might be feeling that way.

In fact, the US is as capable of imposing practical technological and trade-based regulations on Taiwanese companies as it is of regulating technology for Chinese semiconductor companies. It already has a history of regulating the semiconductor industry of its ally Japan from the mid-1980s to the mid-1990s through the first and second US-Japan semiconductor agreements.

Whether it is a matter of production volumes, equipment imports, or even repairs and maintenance on exported equipment, the US controls a large and diverse hand of cards. It is prepared at any moment to wield its authority to permit (or not permit) the use of equipment, materials, components, software, algorithms, or any other technology for which its businesses hold the original patent.

Any of the various economic partnership arrangements or agreements that the US has established with other countries end up shunted down the priority list the moment they are deemed to threaten US interests. This is what allows the US to exert practical control over other countries in every area of advanced industry.

TSMC’s borderline-monopolistic position in the logic semiconductor market is roughly matched by Samsung and SK Hynix’s combined 75% share of the DRAM market. Those companies likewise hold a near oligopoly, and the US certainly does not see overdependence as a desirable situation.

But even with the emergence of DRAM-derived high-bandwidth memory (HBM) and dedicated AI memory semiconductors, the primary market in memory semiconductors is still for general-purpose items, and the US has its own memory semiconductor company in Micron.

Samsung is also building nine new fabs in Texas, a number of which will fabricate memory semiconductors. So its share of memory semiconductor production in the US is likely to remain above a certain level for the time being.

This means that South Korea’s semiconductor industry is relatively safe for now against US curbs — a somewhat different situation from the crisis that TSMC is facing with its monopoly. At the same time, we need to be heedful of the possibility that the US’ attitude toward TSMC as the dominant force in logic semiconductor production could end up also being directed at some point toward Samsung Electronics and SK Hynix.

One aspect of the semiconductor industry restructuring that is in need of discussion is the repurposing of chips for military use.

Combined civilian-military semiconductor technology is something that Washington views as a major issue from a security standpoint in the semiconductor dominance battle between the US and China.

In terms of technological principles, there is no clear distinction between “military” and “civilian” chips. To be sure, those that are intended for military use need to undergo much testing for military environments that are much harsher than those associated with ordinary use. Stable reliability and performance retention take on greater importance than transistor integration density or calculation capabilities.

As a rule, though, semiconductor chips are one of the leading examples of a technology that serves both civilian and military purposes.

When considering the potential for semiconductors to be used for military purposes, there are a few questions we need to analyze. One is why the People’s Liberation Army (PLA) in China has until recently been the biggest institutional importer of Nvidia GPUs.

Another is why the PLA wants high-performance chips that are capable of protection against radiation. A third is why it is using GPU parallel computing to construct environments for simulations of experimental high-temperature, high-pressure conditions.

A fourth question is how GPU-based AI accelerators are being applied to self-operating military drones, unmanned aerial vehicles and urban air mobility. A fifth is how they are being put to use in advanced missile control systems (for flight and other forms of control).

The first three questions can be understood narrowly as examples of seeking advanced semiconductors that are intended for use in nuclear facilities from the development stages of military AI systems. The fourth and fifth can be seen as examples of seeking performance upgrades for unmanned weapon systems and precision strike weapons.

Both mean that a central focus of Chinese semiconductor industry investment is increasing military capabilities — especially attack-related ones.

Things are moving in a direction of increased military tensions in Taiwan and the rest of East Asia and, in the long term, one of reduced US influence in the Asia-Pacific region. That leaves the US in a position where it has to watch the Chinese semiconductor industry ever more closely and apply more intensive restraints.

High-tech industries, including semiconductors, have now grown too much in influence for us to discuss them solely as one economic category in a given country. They are inextricably tied to issues of both domestic and international politics.

We’re entering an era of “techno-politics” as opposed to geopolitics, yet it appears that attitudes toward high-tech industries are still stuck in the frames of industry boundaries and economic spheres of influence.

We’ve been remiss when it comes to analyzing the connections among industries and between those industries and global issues. We need to recognize the fact that South Korea’s “chip war” is not a South Korean issue alone — it’s part of a larger global chip war.

The year 2024 is poised to be an even more crucial moment in light of the global battle for high-tech dominance that is unfolding outside of South Korea. The calendar includes political events such as the US presidential election and general elections in Taiwan and South Korea, while China’s pursuit of semiconductor technology independence will only intensify.

South Korea’s strategic points are in a more three-dimensional situation than ever before. They will need to be objectively redeveloped to suit the aims of acquiring technology in a way that prioritizes economic security.

By Kwon Seok-joon, professor of chemical engineering at Sungkyunkwan University

Please direct questions or comments to [english@hani.co.kr]