
On the day that the stock price of NVIDIA, which has been called the myth of the semiconductor industry, plummeted overnight, we witnessed the paradox of technological innovation. This happened on January 27, 2025, when an AI model created by the Chinese startup DeepSeek was released to the world. The stock market sometimes seems irrational, but that day’s crash may have been a moment that revealed the truth we had been missing.
What made DeepSeek’s emergence so special was that its achievements completely overturned conventional wisdom. The fact that AI models with performance comparable to ChatGPT could be developed for only $6 million shattered the industry’s belief that hundreds of millions of dollars in investment were essential. It was like discovering that a dish that was thought to require expensive ingredients and a long time to prepare could actually be made with a few simple tricks.
This shock is more than a simple technological innovation. The fact that NVIDIA’s market capitalization evaporated 846 trillion won in a single day shows how quickly the market can embrace a paradigm shift. Like the walls of an empire crumbling with a single blow, NVIDIA’s technological advantage and market dominance were shattered in an instant.
Even more interesting is the impact of this incident on the competition for technological hegemony between the U.S. and China. While the U.S. built a high barrier of semiconductor export controls, China found a detour of algorithm efficiency. This seemed to show the law of nature that technological innovation always finds the most efficient way, just as water flows from high to low.
China’s strategy was surprisingly clever. It overcame the limitations of physical resources like GPUs with software innovation. It achieved high performance with fewer computing resources by using a mixed-expert structure, and gained support from developers around the world by releasing it as open source. It was like a chef’s strategy of using an excellent recipe instead of expensive ingredients.
Behind the success of Deepseark, however, there was a dark shadow that we should pay attention to. The results of the Associated Press’ in-depth review were shocking. When asked about the character of Winnie the Pooh, Deepseark gave the superficial answer that it was “a cartoon character loved by Chinese children and families. When asked about the Tiananmen Square protests of 1989, he dodged the question by saying it was “out of scope. When it came to the Taiwan issue, he merely repeated the simple political rhetoric that “Taiwan has been part of China since ancient times.
What was even more disturbing was the severe hallucination phenomenon that DeepSick exhibited. Problems such as talking about something that never happened as if it did, or citing research that did not exist were common. This was like a chef with outstanding skills who sometimes claims to hallucinate and cook with ingredients that do not actually exist. This is a lesson that accuracy and reliability are as important as the efficiency of the technology.
The change in industrial structure brought about by DeepSeeker is particularly noteworthy. The existing method of AI development based on expensive hardware has been fundamentally challenged. This has important implications for the Korean AI industry. As the strategy of focusing on securing GPUs has shown its limitations, the stock prices of software companies such as Naver and Kakao have risen, creating a paradoxical phenomenon.
This change requires a deep reflection on the future of the AI industry in Korea. It is time to move away from the hardware-centric development paradigm and focus more on software optimization and algorithm innovation. In particular, the strategy of developing efficient AI models in specialized fields such as manufacturing, medical care, and finance has become more important.
Amid these changes, Korean startups are also discovering new opportunities. Low-power, high-efficiency NPUs (neural network processing units) developed by companies such as Furiosa and Rebelion will become even more important in the future. As the demand for edge computing is expected to increase with the proliferation of generative AI agents in 2025, these technological innovations can open up new markets.
The government is also keeping pace with these changes. It is preparing active support measures, such as expanding the AI budget to 1.8 trillion won in 2025 and promoting the establishment of a National AI Computing Center. It is also working to reduce its dependence on China by strengthening semiconductor supply chain cooperation with the United States and Japan.
However, these changes also bring new challenges. Issues of personal information protection and AI ethics are becoming more important. The AI Basic Law to be implemented in 2026 reflects these concerns, and the establishment of an ethical AI development system is becoming an urgent task.
We are now at a new inflection point. The emergence of DeepSeeker has brought about a paradigm shift in the entire AI industry that goes beyond mere technological innovation. At the same time, however, it has left a profound lesson that technological innovation does not necessarily lead to the development of truth and reliability.
We now have to ask ourselves new questions. What is real technological innovation? Is it enough to simply find a more efficient and cheaper way? Or should we strive for a higher level of innovation that includes accuracy, authenticity, and ethical values? The innovation and limitations demonstrated by Deep Sea ask us these fundamental questions. Now is the time for us to find the answers and forge our own new path.
Leave a Comment