▲ AI PRISM* Personalized Economic Briefing
* Editor's note: AI PRISM (Personalized Report & Insight Summarizing Media) is an AI-based personalized news recommendation and summary service developed with support from the Korea Press Foundation. It selects and provides six tailored news items by reader type.
[Key Issue Briefing]
■ China's AI Counterattack: Chinese artificial intelligence, long dismissed as offering merely low-cost, value-for-money models, is shaking up the market as competition shifts to technical sophistication. Over roughly 70 days since June this year, nine major Chinese AI firms have poured out 10 latest models and major updates, closing in on top-tier U.S. models in key areas such as reasoning, coding and agents.
■ The Testing-Ground Gap: Korea's AI model development capability ranks among the world's top three, but the country lacks the testing grounds to grow the models it builds into real services and industries. In the second half of last year, Korea's AI adoption rate reached 30.7%, ranking 18th among 147 countries, with the fastest pace of increase. Yet the growing demand has not fully translated into growth for domestic models and services.
■ Memory Super-Cycle: As the memory super-cycle driven by AI expansion continues, the coffers of listed domestic chipmakers are filling up fast. The cash holdings of Samsung Electronics (005930.KS) and SK hynix (000660.KS) have surged by about 117 trillion won this year alone, and the warmth is spreading across the semiconductor materials, parts and equipment sector as a whole.
[Corporate CEO News of Interest]
1. One New Model a Week: Chinese AI Overtakes Big Tech
- Key summary: Chinese AI models, once regarded as low-cost, value-for-money offerings, are shaking up the market as they evolve into a new stage of competing on technical sophistication. Over roughly 70 days since June this year, nine major Chinese AI firms have released 10 latest models and major updates. Starting with Zhipu AI's GLM-5.2, they came in quick succession — Meituan's LongCat-2.0, Tencent's Hy3, Moonshot AI's Kimi K3 and Alibaba's Qwen3.8-Max — while DeepSeek unveiled the official version of V4 Pro on the 13th. On the Artificial Analysis Intelligence Index v4.1.1, DeepSeek V4 Pro 0813 scored 53 on the AI intelligence index, trailing GPT-5.6 Sol (61) and Claude Opus 5 (63), but its average cost per task was $0.06, one-39th the level of Claude Opus 5 ($2.34). Kwon Young-hwan, a principal researcher at the Software Policy & Research Institute, assessed that Chinese firms' influence is spreading rapidly across the open-source AI ecosystem, where lock-in effects arise.
2. Ranked 3rd in AI Model Development, Korea Lacks a Testing Ground
- Key summary: While Korea's AI model development capability has climbed to among the world's top three, critics say the country lacks the testing grounds to deploy and grow these models in real services and industries. As of the end of 2025, notable AI models numbered 59 in the U.S., 35 in China and 8 in Korea, with models from the LG (003550.KS) AI Research, Naver Cloud and SK Telecom (017670.KS) meeting global standards. In addition, Korea's AI adoption rate in the second half of last year was 30.7%, ranking 18th among 147 countries with the fastest pace of increase among surveyed countries. Yet in July this year, monthly users of domestic AI chatbot apps stood at just 1.15 million for A. and 610,000 for Wrtn, compared with 23.67 million for ChatGPT and 9.47 million for Google Gemini. Private AI investment last year was about $1.8 billion in Korea, one-seventh of China's level, against $285.9 billion in the U.S. and $12.4 billion in China. Kang Jun-young, a professor at the Graduate School of International and Area Studies at Hankuk University of Foreign Studies, stressed the importance of a strategy to control chokepoints indispensable to the AI ecosystem, much like memory chips.
3. Amid the Memory Super-Cycle, Samsung-SK hynix War Chest Surges 117 Trillion Won This Year
- Key summary: The cash holdings of Samsung Electronics and SK hynix have surged by about 117 trillion won this year, raising expectations for large-scale capital investment, shareholder returns and merger-and-acquisition (M&A) funding. The two companies' cash holdings and short-term financial products in the second quarter came to 189.999 trillion won and 87.96 trillion won respectively, up 50.97% and 151.75% from the end of last year. According to FnGuide, the cash holdings of 138 listed semiconductor companies in the first quarter of this year totaled 103.46 trillion won, up 26.7% from the end of last year, while the growth rate for all listed companies stood at just 2.4%. Samsung Electronics posted second-quarter revenue and operating profit of 171.5 trillion won and 89.5 trillion won, while SK hynix recorded 79.32 trillion won and 60.54 trillion won. SK hynix projected its investment this year at the high 40 trillion won range.
[Corporate CEO Reference News]
4. Won Rises 9.4%, Biggest Appreciation Among Major Economies
- Key summary: From last month through recent days, the value of the won has risen more than any other major currency. According to the "International Finance and Foreign Exchange Market Trends Since July" released by the Bank of Korea on the 13th, the won appreciated 9.4% against the U.S. dollar from the 1st of last month to the 11th of this month, with the won-dollar exchange rate falling from 1,549.4 won at the end of June to 1,416.0 won on the 11th of this month. The BOK explained that the dollar weakened amid eased uncertainty in the Middle East and weaker expectations for U.S. rate hikes, while inflows from SK hynix's conversion of funds raised through its U.S. American Depositary Receipt (ADR) issuance also served as a factor behind the won's strength. Meanwhile, the average daily change in the won-dollar exchange rate in July was 0.53%, the second-highest after the Russian ruble (0.55%), meaning volatility grew as well.
5. Chung Ki-sun, Stressing a Long-Term View: "Prepare Fiercely for Future Businesses"
- Key summary: HD Hyundai (267250.KS) Chairman Chung Ki-sun urged the group's presidents to pursue sustainable growth rather than short-term results, calling for future-oriented leadership. At a group presidents' meeting held at the HD Hyundai Global R&D Center in Pangyo — attended by Vice Chairmen Cho Young-cheol, Cho Seok and Lee Sang-kyun along with the CEOs of 11 affiliates spanning shipbuilding and offshore, energy and construction equipment — Chung referred to the painful experience of failing to read market changes because of an obsession with immediate results. The meeting reviewed each company's first-half results, new businesses and future growth strategies, and shared group-wide priorities such as accelerating AI transformation (AX). HD Hyundai achieved cumulative first-half revenue of 42.0113 trillion won and operating profit of 6.9594 trillion won on balanced growth across major businesses including shipbuilding, construction equipment, refining and power equipment. Chung said he would create a management environment to lead sustainable growth with a long-term view of 10 to 20 years.
6. Naver Bets on "AIDC at Sea," Strengthening Its AI Infrastructure Competitiveness
- Key summary: Naver has invested in offshore AI data center technology that generates its own power from wave energy, moving to strengthen its AI infrastructure competitiveness. Naver said on the 13th that it invested in Pantalassa, an offshore AI data center startup, through the U.S. venture capital firm 8090 Industries. Founded in 2016, Pantalassa is developing a floating node platform that generates power from wave energy without a power plant or fuel and uses it directly for AI computation. The system cools servers with seawater and transmits only the resulting data to the ground via low-Earth-orbit (LEO) satellite communication networks such as Starlink, reducing reliance on transmission grids and undersea cables. Naver CEO Choi Soo-yeon said the company would proactively invest in next-generation technologies such as wave-based renewable-energy AI data centers to strengthen its competitiveness in multiple ways in the global AI infrastructure market.
