South Korea's Espionage Law Won't Stop What's Already Happening in DRAM

South Korea took its espionage law — unchanged since the Cold War — and on September 13 expanded it to cover any foreign country, not just North Korea. Courts can now sentence people who pass sensitive technology abroad to up to 30 years in prison. The legislation was framed as armor for the nation's semiconductor industry, home to Samsung Electronics and SK Hynix, the world's two largest DRAM producers.

The law looks like a wall. It's not. It addresses a threat that's already been overtaken by a different one — one that no criminal statute can stop.

The leak that started it

The push for reform follows a pattern of industrial technology theft that reads like a corporate security nightmare. In December, prosecutors indicted 10 former Samsung employees for leaking 18-nanometer DRAM process technology — the manufacturing recipe containing roughly 600 production steps — to China's ChangXin Memory Technologies, known as CXMT. Samsung spent five years and 1.6 trillion won ($1.1 billion) developing that technology. South Korean police estimated the leak cost Samsung roughly 5 trillion won, or $3.7 billion, in lost sales during 2024.

This wasn't an isolated incident. A former Samsung team manager already received a seven-year prison sentence for leaking the same technology earlier. The pattern was systematic: CXMT, founded in 2016 with no research facilities or manufacturing equipment, recruited Samsung veterans who then adapted the stolen "Process Recipe Plan" for Chinese manufacturing tools. By 2023, CXMT was mass-producing 10-nanometer-class DRAM chips.

The espionage law was a response to a real problem. But the problem has moved on.

The numbers the law doesn't change

In the second quarter of 2026, CXMT captured 10% of the global DRAM market. A year earlier, it had 4%.. That is a 150% increase in global market position in one year, for a company that existed as a credible competitor only a few years ago. The same quarter saw SK Hynix's standard DRAM share fall from 39% to 25%.

CXMT's growth wasn't driven solely by stolen know-how. The company just completed an $8.6 billion IPO on the Shanghai stock exchange, with shares jumping 466% on the first day of trading. It has received 2.6 trillion won in Chinese government investment. It can now produce up to 280,000 wafers per month.. And it's using that capital to target HBM — the high-bandwidth memory that powers AI servers — and LPDDR6, the next generation of mobile memory.

The DRAM market is booming — global demand surged 385% year-over-year in Q2 2026, driven by AI servers and next-generation GPUs. But CXMT's revenue grew 716% year-over-year, more than double the market rate. That is not a leak problem. That is a subsidy-and-capacity problem.

What the law can't prosecute

Here's the mechanical gap between the headline and the reality. A professor at Seokyung University identified what legal experts call a "huge blind spot" in the new framework: prosecutors must prove nation-state involvement to apply espionage charges. The information must be classified as a "national secret," and there must be a demonstrable connection between the accused and a foreign government.

China's own National Intelligence Law requires all Chinese citizens and organizations to support and assist state intelligence work. That makes it legally impossible for a Chinese national to deny acting on behalf of the state, even when they're motivated by private commercial gain. The result is a catch-22: the very mechanism that enables the theft also makes it nearly impossible to prove under the new law's stricter standard.

Leaks that don't meet the espionage threshold still fall under the Industrial Technology Protection Act, which carries its own penalties. But those penalties are the ones the new law was meant to replace — the ones considered too light to be a deterrent.

The stronger the penalties on paper, the more important it becomes that prosecutions actually happen. And the prosecution bar just got higher.

The bigger incumbent failure

The espionage law focuses investor attention on theft, but there's a more immediate reason SK Hynix is losing ground — one that has nothing to do with spies.

SK Hynix signed long-term agreements with customers at fixed prices before the DRAM market's current price surge. Those locked-in prices are now well below what the open market commands. SK Hynix's HBM market share fell from 64% in Q2 2025 to 50% in Q2 2026, despite revenue surging 214% year-over-year. The share loss reflects the mathematics of pricing, not volume: SK Hynix sold more chips but at lower average prices, shrinking its revenue share even as the market exploded.

Samsung avoided this trap. It didn't sign LTAs as aggressively, rode the price surge, and reclaimed the #1 standard DRAM position at 38% share. SK Hynix isn't just being undercut by a Chinese rival — it's being hurt by its own deal structure.

This is the Graviton pattern in a different dress: a challenger gains share partly because the incumbent created its own opening. The difference here is that the opening isn't technological incompetence — it's pricing discipline. SK Hynix gave away margin to lock in customers, and the market moved away from it.

What this means for investors

The espionage law is political theater with real legislative backing. It signals that South Korea treats its semiconductor industry as a national security asset, which it is — semiconductors account for roughly 20% of South Korea's exports. But it doesn't change the competitive arithmetic.

For the three publicly traded memory producers — Samsung and SK Hynix in South Korea, Micron in the U.S. — the question isn't whether Seoul passed a law. It's whether CXMT's capital, capacity, and government backing can sustain a long-term push into higher-margin products like HBM. The company's $8.6 billion IPO specifically targets that ambition.

The DRAM market is currently in a high-price, high-margin cycle. Micron's gross margins sit at 72.6%, with revenue growth of 167% year-over-year. These are extraordinary numbers, but memory is a commodity cycle business. High prices attract capacity. CXMT is building exactly that capacity, backed by capital that doesn't need to justify returns to public shareholders in the same way.

The law won't stop CXMT. Capacity and pricing will. And on that battleground, criminal statutes don't matter.