Tech Explained: Alliances, policy and memory supercycle define Korea’s AI push in 2025 in Simple Terms

Korea’s artificial intelligence (AI) push met a decisive moment in 2025 as global alliances, national policy and hardware demand aligned to create noteworthy momentum.

Nvidia and OpenAI deepened cooperation with Korea, unveiling plans for large-scale AI infrastructure backed by the government and local conglomerates. At the same time, the government elevated sovereign AI to a core national strategy by sharply increasing budgets, and the National Assembly passed a comprehensive AI framework law set to take effect next year, making Korea the first country to implement such policies.

Companies have moved beyond developing language models to expand real-world applications of AI, while the hardware sector is enjoying a memory supercycle driven by surging data center demand. The boom has created supply bottlenecks not only in AI-specific high-bandwidth memory (HBM) but also in conventional memory semiconductors.

President Lee Jae Myung poses with attendees of a meeting with OpenAI CEO Sam Altman, fourth from left, at the presidential office in Seoul, Oct. 1. From third from left are Samsung Electronics Executive Chairman Lee Jae-yong, Altman, President Lee and SK Group Chairman Chey Tae-won. Courtesy of presidential office

Deepening ties with global AI giants

Nvidia is positioned at the center of Korea’s efforts to secure sufficient AI accelerators. In October, Nvidia CEO Jensen Huang visited Korea for 2025 Asia-Pacific Economic Cooperation meetings, and announced a plan to supply 260,000 Nvidia Blackwell graphics processing units (GPUs) for the government, Samsung Electronics, SK Group, Hyundai Motor Group and Naver.

With the Nvidia GPUs, Samsung, SK and Hyundai Motor plan to empower their manufacturing capabilities with AI factories and digital twin technologies. Digital twin is a technology mirroring real-world entities in a virtual environment. Naver will expand the capacity of its hyperscale data centers while enhancing physical AI applications. The government plans to deploy the GPUs to support AI development across companies and industries, develop sovereign AI foundation models and build a national computing center.

OpenAI CEO Sam Altman also visited Korea multiple times this year, which resulted in Samsung and SK joining the Stargate project, a $500 billion joint venture for AI infrastructure in the U.S. As part of their partnerships, Samsung Electronics and SK hynix will each supply memory chips to meet OpenAI’s chip demands — about 900,000 wafers’ worth of high-performance dynamic random-access memory (DRAM) each month.

The deepening ties with Nvidia and OpenAI mark a turning point for Korea’s AI strategy. By anchoring its AI ambitions in partnerships with leading global players, Korea is moving more decisively to secure AI infrastructure competitiveness amid intensifying global competition.

AI Basic Law

In January, Korea enacted the Framework Act on the Development of AI and the Creation of a Foundation for Trust, which is set to take effect on Jan. 22 next year, making the nation the first in the world to enact such a policy.

Designed to promote and regulate the domestic AI industry, the act defines high-risk AI systems; introduces obligations related to safety, transparency and accountability; and establishes a governance framework covering areas such as data use, human oversight and liability.

The act, widely called the AI Basic Law, is expected to provide regulatory clarity for companies developing or deploying AI systems in Korea. This is expected to help build trust in AI technologies, reduce legal uncertainty and accelerate AI adoption across industries by setting clear rules on responsibility and risk management.

However, concerns have been raised over unclear definitions of high-risk AI, the difficulty smaller businesses and startups may face in complying with the rules, and ambiguity over labeling requirements for AI-generated content. Industry officials warn that such uncertainties could discourage innovation or slow commercialization, particularly during the early stages of implementation.

In response, the government is now making efforts to clarify those standards through enforcement decrees and by coordinating between industry, government and civic groups. Authorities are also considering grace periods for penalties, including administrative fines, to allow companies time to adapt once the law takes effect.

Members of the Presidential Council on National Artificial Intelligence Strategy cut a cake in celebration of the first 100 days of the council’s launch in Seoul, Dec. 15. Yonhap

Sovereign AI

An emblem for companies participating in the government’s project to develop Korea’s proprietary artificial intelligence foundation model / Courtesy of Ministry of Science and ICT

With the start of the Lee Jae Myung administration in June, Korea has identified AI as a top policy priority, setting an ambitious goal of becoming one of the world’s top three AI powers based on AI sovereignty. Lee appointed AI expert Bae Kyung-hoon as minister of science and ICT and elevated the post to the level of deputy prime minister, granting broader authority over AI, semiconductors and digital policies.

The president also launched the Presidential Council on National AI Strategy to oversee the government’s plans to build a belt of national AI computing centers, power generation facilities and related infrastructure, while fostering growth in businesses and industries focused on neural processing units and non-GPU AI accelerators.

As part of the push for AI sovereignty, the government earmarked 10.1 trillion won ($6.94 billion) in the national budget for AI, roughly tripling this year’s level, and launched a project to develop a proprietary foundation model through a competitive program involving the country’s leading AI model developers.

Through these measures, the government seeks to position AI as a strategic national asset, while lowering the country’s reliance on foreign tech companies.

A robotic arm picks up an object during the 2025 Robot World exhibition at KINTEX in Goyang, Gyeonggi Province, Nov. 5. Yonhap

Real-world applications and agentic AI

The domestic AI industry has experienced a rapid shift in focus this year, moving away from developing large or small language models and toward applying AI to real-world use cases.

Leading conglomerates have already introduced AI chatbots for internal knowledge management and customer response, while automating office document processing, data analysis and demand forecasting across their operations.

AI is also increasingly being used in advertising and content production, while advances in translation and interpretation technologies are enabling companies to expand services for global customers. Although still at an early stage, agentic AI systems are also emerging, executing various tasks on behalf of users such as booking restaurants or hotels and completing payments.

In manufacturing, physical AI — where intelligence is embedded directly into production systems — has become one of the most representative application areas, with AI deployed for real-time process anomaly detection and automated quality inspection on production lines. Combined with advances in robotics, companies are accelerating efforts toward broader manufacturing automation.

A high-bandwidth memory 4 chip is displayed at SK hynix’s booth at the SEDEX 2025 exhibition in Seoul, Oct. 22. Yonhap

Memory supercycle

The rapid expansion in AI services resulted in exponential growth of demand for data centers powered by AI accelerators and memory chips, triggering a supercycle.

The uptrend has been particularly centered on HBM, with memory chip makers such as Samsung Electronics and SK hynix struggling to keep pace with demand even while running production at full capacity.

This supply bottleneck has pushed HBM prices higher, with the companies reportedly raising the supply prices of HBM3E chips by about 20 percent. Given that the two memory giants are set to begin mass production of next-generation HBM4 next year, the fact that prices for current generation products remain high is considered unusual.

As memory makers concentrate their manufacturing capacity on HBMs, prices of general-purpose DRAM products have also risen. According to market tracker TrendForce, contract prices for DRAM surged about 420 percent this year, jumping from $3.75 in January to $19.50 in November.

The impact is expected to spill over into the prices of other computing devices, such as smartphones.

Analysts are almost certain that both Samsung Electronics and SK hynix will enjoy solid earnings in coming months. Meanwhile, the market’s attention is now shifting toward how the AI trend will benefit other chips, such as low-power memory, application-specific integrated circuits, system semiconductors and NAND flash memory.