[Editorial] China’s Rapid AI Advance Shows South Korea Cannot Rely on Semiconductors Alone to Secure Leadership

The global AI industry is being shaken by Kimi K3, the new model recently unveiled by Chinese AI startup Moonshot AI. Kimi K3 is a massive AI model with 280 billion parameters and a context window of 1 million tokens. Artificial Analysis, an independent AI model evaluation firm, gave Kimi K3 an intelligence score of 57. That puts it close behind Anthropic's Claude Fable 5, which scored 60 and is regarded as the best-performing model available, and OpenAI's GPT-5.6 Sol, which scored 59. Chinese AI, which has grown despite U.S. sanctions on high-performance semiconductors, is now moving beyond price competition and closing in on the United States in top-tier performance as well. The world has no choice but to watch the ripple effects.
Moonshot was also behind the weekend plunge in global stock markets. U.S. tech shares fell sharply as doubts grew over whether American companies, after pouring in astronomical investment, can generate returns that match expectations. The selloff spread as skepticism mounted over Big Tech's expensive closed models and the enormous semiconductor demand that supports them. It can be seen as the second warning after last year's DeepSeek shock from China.
Of course, it is still too early to say that U.S. AI competitiveness has been overtaken by China. OpenAI and Anthropic continue to lead the race for top-tier models. Still, Google has delayed the launch of its next-generation Gemini because of performance improvement issues, showing that U.S. Big Tech is also spending considerable time and money on technical upgrades. Chinese companies are exploiting this gap and rapidly narrowing the distance. In particular, they are widening choices for developers and companies around the world with open models, while also putting cracks in the U.S.-centered AI ecosystem.
We cannot avoid looking at our own situation, where the gap with the United States and China is not narrowing but widening. South Korea has world-class competitiveness in High Bandwidth Memory and advanced semiconductor production. But AI leadership is not determined by semiconductors alone. Foundation models, cloud, data, software, platforms, developer ecosystems, and global services must all work together.
The government's Sovereign AI project is the minimum line of defense for building an independent ecosystem. The initiative began last year with five elite teams and, after stage evaluations, is now moving forward as a four-way competition among LG AI Research, SK Telecom Co., Ltd. (SKT), Upstage, and the newly selected Motif Technologies. The government says it has secured 35,000 of the 50,000 advanced Graphics Processing Units (GPU) it targeted for the first half of this year and laid the groundwork for an independent model that could rank among the world's top three. It is encouraging that the effort has not stood still, but given the pace of the United States and China, it is hard to say the results or speed are satisfactory.
It is also urgent to bring in genius-level talent and build a stable training system. South Korea should use the United States' tighter rules on foreign students and researchers as an opportunity to attract global talent. The United States abruptly announced measures that will limit the stay of foreign students and researchers to a maximum of four years starting in September. We suggest carefully designing and implementing talent recruitment measures aimed at the world's master's and doctoral candidates, as well as researchers, in advanced fields.
The Kimi K3 frenzy is a reminder that AI leadership is not the exclusive property of the United States. At the same time, South Korea must take to heart the warning that its future is not guaranteed simply by its current status as a semiconductor powerhouse.