As artificial intelligence (AI) semiconductors gain prominence, South Korean companies are standing out in the field of neural processing units (NPUs). However, there are growing fears about technology and talent being siphoned off to China. Reports indicate that Chinese semiconductor companies are actively recruiting S. Korean engineers and trying to acquire critical technologies.
According to industry sources some S. Korean AI chip design companies (known as fabless firms) and intellectual property (IP) developers are stepping up internal security to prevent technology leaks to China. While China is a major market for S. Korean semiconductor exports, it is also a competitor. U.S. restrictions on semiconductor exports to China have made it difficult for Chinese companies to access advanced AI chips.
China, in particular, has struggled to import cutting-edge GPUs (graphics processing units) from Nvidia for its AI data centers. With limited access, Chinese firms have resorted to using lower-performance GPUs. Some have started developing their own NPUs to fill the gap, but industry experts note that these efforts lag behind in both performance and commercialization.
As a result, S. Korea’s rapid advancements in NPU technology have made its companies a prime target for Chinese interests.
According to a representative from S. Korea’s largest AI fabless company, many local NPU firms have reached a stage where their AI chip development can rival GPUs. These companies are also collaborating with foundries to improve yield rates and performance. “Aside from the United States and Israel, S. Korea is leading the race in NPU development,” the representative said.
This interest from China is not new. Chinese companies have previously acquired NPU-related IP from S. Korea.
In Sept. 2024, a Chinese AI semiconductor company signed a licensing deal worth $500 million with Seoul-based IP provider Chips&Media. Before that, OpenEdges, another S. Korean IP developer, entered a licensing agreement with a Chinese firm.
The growing appeal of NPU technology lies in its potential to power not only AI chips for data centers but also a wide range of devices. According to market research firm IDC, by 2026, half of all IT devices sold in China are expected to feature AI engines, including NPUs. This projection highlights the increasing role of NPU technology in enabling AI functionality across diverse applications. Currently, S. Korea, the United States, and Israel dominate the NPU field, with S. Korea leading in development speed, according to analysts.
Despite these advancements, S. Korean NPU firms face challenges in safeguarding their technologies. While the government has implemented legal frameworks to protect memory chip technologies, mid-sized and startup NPU companies often operate in a regulatory blind spot. Industry experts are calling for legislative and regulatory measures to better protect these emerging technologies.
The risks of technology leakage to China remain significant. Recent incidents involving S. Korean memory semiconductor technology have heightened industry vigilance.
For instance, a former Samsung Electronics employee is on trial for allegedly leaking 18-nanometer DRAM processing information to Chinese company ChangXin Memory Technologies (CXMT) in 2016. In response, South Korea’s National Assembly recently passed amendments to the Industrial Technology Protection Act, increasing penalties for leaks of core national technologies.
