Semiconductors, Sovereign Compute, and Strategic Ambition
By Javier
Surasky
.png)
South Korea
occupies an intermediate position on the artificial intelligence power map: it
neither controls the leading foundation models nor possesses a global cloud
platform comparable to those of the United States, and therefore depends on
foreign technologies to access some of the most advanced layers of computing
(Richardson et al., 2025).
At the same
time, the country has advanced industrial capabilities in semiconductor
manufacturing, particularly in cutting-edge memory, an area in which two South
Korean companies stand out: Samsung Electronics and SK Hynix. Together, they
make South Korea an industrial supplier embedded in a sensitive global chain
that controls the critical material layers underpinning AI (Ministry of Science
and ICT [MSIT], 2025a).
Samsung
Electronics brings together capabilities in memory, foundry services, design,
consumer electronics, mobile devices, and data centers. Its scale and vertical
integration allow it to operate across different layers of the digital
ecosystem, where it faces competition from a giant such as TSMC and from U.S.
companies that dominate accelerator design and AI software (MSIT, 2025a).
SK hynix is
a leading and fast-growing player, thanks to its strong position in the
high-bandwidth memory (HBM) market, and it maintains a close relationship with Nvidia, which needs its memory products to manufacture accelerators (NVIDIA,
2025).
This
combination of strengths and dependence explains its current strategic
ambition: to transform its industrial leadership in semiconductors into a broad
capacity to influence global decisions on infrastructure, computing, models,
data, standards, and regulation. This techno-state orientation seeks to achieve
“technological sovereignty to increase self-sufficiency in advanced
technologies” (Lee, 2025, p. 484), reduce vulnerabilities, secure access to
critical resources, and strengthen its bargaining power.
The
challenge is to move from being a component power to becoming a systems power
as well, gaining greater influence over global AI decisions.
In this
process, South Korea is caught in a political tension: it is a U.S. security ally, while Samsung and SK Hynix operate plants in China. This situation has
been described as a “sandwich dilemma” (Moon & Yeon, 2025, p. 549) and runs
throughout the technology policy defined in Seoul.
Its
strategy combines financing, tax incentives, talent development, research,
industrial policy, public procurement, infrastructure construction, and
coordination among national public and private actors. One example is the
semiconductor mega-cluster project in Gyeonggi Province, which seeks to connect
production facilities, suppliers, research centers, materials and equipment
companies, and training facilities around areas such as Yongin, Pyeongtaek,
Hwaseong, Icheon, Suwon, and Pangyo, as part of a state-planning exercise
extending through 2047 (MSIT, 2025a).
South Korea
is also working to expand its sovereign computing capacity, treating access to
graphics processing units and data centers as a prerequisite for sustaining
domestic AI development. In February 2025, it announced a target of securing
10,000 high-performance GPUs for a national computing infrastructure. Some time
later, the government, Nvidia, and major business conglomerates presented plans
to deploy more than a quarter of a million GPUs across sovereign clouds, public
centers, and private facilities, turning actors such as Samsung itself, SK
Group, Hyundai Motor Group, Naver Cloud, and Kakao into critical players
(NVIDIA, 2025; Reuters, 2025).
Computing
has thus become a strategic infrastructure in South Korea, comparable to energy
or telecommunications. Its development, however, is not being pursued at any
cost or by any means: the apparent premise is to avoid foreign hardware and
software, as well as dependence on foreign licenses that would increase the
country’s reliance on external actors (Richardson et al., 2025).
This issue
is especially important for startups, universities, and small and medium-sized enterprises, which face high costs in accessing computing capacity.
South Korea
is also deliberately seeking to develop its own accelerators in order to reduce
its national dependence on Nvidia and become an alternative to it. Companies
such as Rebellions, FuriosaAI, and DeepX show its determination to move from
memory production into AI chip design (Reuters, 2023).
The
situation is similar in the cloud services sector: domestic actors such as SK
Telecom, KT, Naver Cloud, and Kakao are seeking to compete with market
heavyweights such as AWS, Microsoft Azure, and Google Cloud.
In every
case, the development of national models has an economic and political
dimension connected to South Korea’s own language and culture. This opens
another window: the development of sovereign models is intended not only to
compete with foreign counterparts but also to ensure adequate representation of
the Korean language and culture while strengthening sovereign decision-making
capacity in data management.
These
industrial and material dimensions are given legal expression in the Framework
Act on the Development of Artificial Intelligence and the Creation of a
Foundation for Trust, passed by the National Assembly in December 2024,
promulgated in January 2025, and in force since January 2026. The Act
establishes a triangular system whose central pillars are promoting industry,
organizing public AI governance, and addressing the most serious risks arising
from the use of AI. It is therefore both an instrument for imposing
restrictions and a means of advancing industrial policy (MSIT, 2026a).
Among other
categories, the law distinguishes between generative AI and so-called
high-impact AI, which includes systems used in areas capable of significantly
affecting people’s lives, safety, or rights, such as health care,
transportation, and employment. Operators of these systems must adopt measures
relating to risk management, documentation, transparency, explainability, and
human oversight (Korea Legislation Research Institute [KLRI], 2026; MSIT,
2025b).
There are
also obligations to disclose when a service uses generative or high-impact AI
and to identify artificial content that could be mistaken for real content, for
which the Ministry of Science and ICT has issued guidelines (MSIT, 2026b).
In line
with the national policy of increasing digital sovereignty, the law may apply
to foreign companies offering services in South Korea. This gives it a degree
of extraterritorial reach, allowing the state to assert its authority over
platforms and providers that have no headquarters in the country but whose
systems affect users, markets, or institutions within it (KLRI, 2026).
This policy
is projected internationally through events such as the 2024 Seoul AI Summit,
which enabled the country to position itself within debates on international AI
governance. In that context, “Seoul’s approach has been cautious and prudent”
(Moon & Yeon, 2025, p. 564), seeking to turn the tension in its relations with the United States and China into an asset by presenting itself as a
reliable partner and an interlocutor capable of understanding both powers.
The true
measure of South Korean power will be its ability to integrate all these
dimensions without creating an internal concentration of power it seeks to
reduce in relation to external actors, or entering into conflict with either of
the two AI superpowers.
Key Facts
South Korea
is a global semiconductor manufacturing power, especially in DRAM, NAND, and
HBM memory (MSIT, 2025a).
Samsung
Electronics and SK Hynix are the central corporate actors in its semiconductor
industry: the former has strong capabilities in memory, foundry services,
design, and devices, while the latter stands out in HBM (NVIDIA, 2025; MSIT,
2025a).
The
Gyeonggi semiconductor mega-cluster project includes new factories, suppliers,
research centers, and industrial infrastructure extending toward 2047, with
planned investments of 622 trillion won, approximately USD 415 billion at the
July 2026 exchange rate (MSIT, 2025a).
In February
2025, South Korea announced its intention to secure 10,000 high-performance
GPUs to develop a national computing infrastructure (Reuters, 2025).
In October
2025, Nvidia announced agreements with the government and major South Korean
companies to deploy more than 250,000 GPUs across sovereign clouds and public
and private AI facilities (NVIDIA, 2025).
Rebellions,
FuriosaAI, and DeepX represent the effort to develop domestic accelerators and
reduce dependence on foreign AI chip manufacturers (Reuters, 2023).
The country
has a Framework Act on the Development of Artificial Intelligence and the
Creation of a Foundation for Trust, in force since January 22, 2026, which
combines industrial promotion, public governance, and safety and transparency
obligations for generative AI, systems with a high social impact, and certain
large-capacity models, defined according to their processing capacity measured
in FLOPs (KLRI, 2026; MSIT, 2025b, 2026a).
The country
is a U.S. security ally and maintains strong relations with China, creating a
potential tension that its foreign policy seeks to turn into an opportunity to
act as a partner capable of facilitating dialogue between the two.
References
Korea
Legislation Research Institute. (2026). Framework Act on the Development of
Artificial Intelligence and the Creation of a Foundation for Trust (Act No.
20676, Jan. 21, 2025, as amended by Act No. 21311, Jan. 20, 2026).
https://elaw.klri.re.kr/eng_service/lawView.do?hseq=73499&lang=ENG
Lee, S.
(2025). High technology and economic statecraft: The emergence of
techno-economic statecraft in South Korea. Business and Politics, 27,
484–503.
https://doi.org/10.1017/bap.2023.34
Ministry of
Science and ICT, Republic of Korea. (2025a). Government to establish the
world’s largest and most advanced mega semiconductor cluster.
https://www.msit.go.kr/eng/bbs/view.do?bbsSeqNo=42&mId=4&nttSeqNo=952&sCode=eng
Ministry of
Science and ICT, Republic of Korea. (2025b). MSIT announces legislative
notice for the Enforcement Decree of the AI Basic Act to foster the AI industry
and build a foundation for safety and trust.
https://www.msit.go.kr/eng/bbs/view.do?bbsSeqNo=42&mId=4&mPid=2&nttSeqNo=1191&sCode=eng
Ministry of
Science and ICT, Republic of Korea. (2026a, January 22). The AI Basic Act
comes into force to lay the foundation for Korea to become an AI G3.
https://www.msit.go.kr/eng/bbs/view.do?bbsSeqNo=42&mId=4&mPid=2&nttSeqNo=1214&sCode=eng
Ministry of
Science and ICT, Republic of Korea. (2026b, January 22). MSIT releases
guidelines on ensuring AI transparency.
https://www.msit.go.kr/eng/bbs/view.do?bbsSeqNo=42&mId=4&nttSeqNo=1215&sCode=eng
Moon,
C.-I., & Yeon, W. (2025). Clashes of techno-statecraft: US-China technology
rivalry and South Korea’s strategy. Business and Politics, 27, 548–568.
https://doi.org/10.1017/bap.2024.26
NVIDIA.
(2025, October 30). NVIDIA, South Korea government and industrial giants
build AI infrastructure and ecosystem to fuel Korea’s next wave of innovation.
https://nvidianews.nvidia.com/news/south-korea-ai-infrastructure
Richardson,
A., Yi, H., Nie, M., Wisdom, S., Price, C., Weijers, R., Veld, S., & Baker,
M. (2025). How sovereign is sovereign compute? A review of 775 non-U.S. data
centers. arXiv.
https://arxiv.org/pdf/2508.00932
Reuters.
(2023, February 13). South Korea aims to join AI race as startup Rebellions
launches new chip.
Reuters.
(2025, February 17). South Korea aims to secure 10,000 GPUs for national AI
computing centre.