China's Changxin Memory Technologies (CXMT) has climbed to the fourth position in the global DRAM market, holding an approximate 8% market share. In the first quarter of this year, the company's revenue surged by 719% compared to the previous year. CXMT reported its first profit since its establishment in 2025 and is now seeking to raise billions through a listing on the Shanghai Stock Exchange. The company has announced that it is mass-producing both server-grade DDR5 and mobile LPDDR5X, while also selling older DDR4 at about half the market price. Just a few years ago, CXMT was considered to be five years behind its competitors.
The semiconductor industry is notoriously difficult to catch up in, as advanced processes are perfected through records of failures rather than blueprints. Understanding where defects recur and under what conditions yield curves decline requires thousands of trials. While equipment can be purchased, the conditions for operating that equipment cannot be bought. The time accumulated over 30 years by South Korean memory manufacturers is the real barrier to entry.
When a latecomer's timeline accelerates, the root cause is often attributed to technology leakage. However, it would be simplistic to attribute China's rapid progress solely to leaks. State subsidies, a massive domestic market, and capital that can absorb losses have all played a role. Nevertheless, if the learning curve has notably shortened, it is reasonable to assume that someone has provided materials that allow for skipping steps in that curve.
Such incidents have occurred repeatedly. A former executive accused of leaking Samsung Electronics' 18-nanometer DRAM process information was sentenced to six years and four months in prison, along with a fine of 200 million won, in a retrial. Another former employee, who allegedly printed process data while moving from SK Hynix to a Huawei subsidiary, received a five-year prison sentence and a fine of 30 million won in an appeals court.
Industry insiders agree that sentences of five to six years do not deter the incentive structure. Chinese companies offer core personnel several times their existing salaries. Once the contract is fulfilled, individuals retain assets that would be difficult to create even after 20 years of work in South Korea. Even if caught by South Korean law, the calculation becomes one of financial gain after serving time. Fortunately, both cases saw heavier sentences due to amendments to the Industrial Technology Protection Act and the Unfair Competition Prevention Act.
The criteria for determining the severity of the crime are also unclear. While the number of files taken, the market value of the data, and the compensation received by the leaker are used to assess damages, the critical factor is not the number of files but the time lost in narrowing the technology gap.
Industrial technology leakage is not merely a job dispute involving individuals who have betrayed their companies. It involves the transfer of assets built over decades through taxpayer investment, human resources, and time to a competing nation, with individuals reaping the rewards. The punishment system must reflect this nature. The sentencing guidelines should treat the overseas leakage of core national technologies as an independent aggravating factor, and the criteria for calculating damages should shift from the compensation received by the leaker to the time and cost required to regain the technology. The confiscation and recovery of criminal proceeds must be enforced to prevent the calculation of 'profitable imprisonment.' Separate penalties for brokers and intermediary organizations facilitating personnel transfers, as well as monitoring of indirect leaks through overseas corporations and partners, are also necessary.
The costs of technology leakage are severe. The general-purpose DRAM market has already seen significant losses. In a market where Chinese companies are selling DDR4 at half the price and have entered the production stage for DDR5, South Korean firms are inevitably experiencing a decline in market share. The remaining defensive line lies in high-bandwidth memory (HBM) and advanced AI semiconductors. Even CXMT acknowledges a 2-4 year gap behind South Korean companies in HBM technology.
That gap can vanish with just a few pieces of data. HBM technology involves a combination of core DRAM, base die, stacking and bonding, and thermal design. While individual processes can be inferred, achieving the right conditions simultaneously can only be gained through experience. The current transition to custom HBM and advanced packaging is the most vulnerable period for leaks.
Once technology has been transferred, it cannot be reclaimed through any ruling. The semiconductor super-gap is determined not by how quickly new products can be made, but by how well the existing 2-4 year lead can be maintained.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.