Seoul's chip-spy law exposes the pressure on the memory duopoly

On September 13th South Korea's revised espionage law took effect, allowing courts to sentence, to as much as 30 years in prison, those who hand sensitive technology to a foreign country or an organisation equivalent to one — a scope that still leaves prosecutors proving that a foreign government directed the leak. On its face the change reads like a response to a scandal: a tightening of the rules on commercial theft. In fact it is a document about the economics of the global memory-chip industry, and about a country losing faith in the very moat that has made its two champions rich.

That moat has, for the first time, become cheap and easy for American investors to bet on. On July 10th SK Hynix, the world's second-largest memory maker and the dominant supplier of the high-bandwidth memory that feeds Nvidia's AI accelerators, listed American depositary receipts on the Nasdaq, raising $26.5 billion. Samsung Electronics, its bigger rival, remains foreign-listed by comparison. A durable Korean duopoly is thus now, in practical terms, a question a United States retail portfolio can own.

Why would a law against spies matter for a company's share price? Because the source of the duopoly's profit — scarce, high-rent memory capacity at the heart of the artificial-intelligence boom — is being attacked where it is weakest. Samsung and SK Hynix together control roughly six in ten DRAM chips sold globally. Chinese rivals are closing the gap with alarming speed: CXMT reached 10% of the global DRAM market in the second quarter of 2026, after an initial public offering that made it China's largest onshore-listed company, and it sells chips at roughly a quarter to a third of the prevailing world price. One leak already did measurable damage: ten former Samsung engineers were indicted for allegedly handing CXMT the production process for 18-nanometre DRAM, costing Samsung an estimated 5 trillion won (about $3.7 billion) in lost sales in 2024 alone.

The law is the institutional answer to that threat, and it exposes the limits of its own instrument. Prosecution still requires proving that a foreign government directed the leak — a requirement that Chae Seong-joon, a former spy-agency official, calls a "huge blind spot", and one that China's system, in which citizens are legally obliged to help the state collect information, is built to exploit. The numbers bear out the weakness. Half of the leaks of sensitive technology detected last year involved China; more than 85% of leaks between 2020 and 2024 hit small and medium-sized suppliers, which lack the guarded fabs and the legal teams on which Samsung and SK Hynix rely. Seoul has armed itself to protect the crown jewels. The perimeter is what is leaking.

To be sure, the statute is not empty theater. It raises the price of betrayal, and it signals to customers, suppliers and capital markets that the state stands behind the incumbents' intellectual property. The trouble is that a law aimed at individual leakers does little against a state whose strategy is to build the capacity itself, at subsidised prices, and to outlast export controls. CXMT, blacklisted by the Pentagon in February and accused of violating American export rules with its 18-nanometre memory, is suing the Pentagon; Apple is among the customers lobbying to buy its chips anyway. The leak channel that matters most is not a rogue engineer in an airport lounge but a national industrial policy that treats acquisition of this technology as a strategic objective.

This is where the story turns from trade policy into valuation. The memory supercycle's rents are real: SK Hynix's earnings have climbed roughly an order of magnitude through the AI buildout, and surveyors of analyst opinion label the shares a buy. But an earnings spike is not a moat. What investors in SK Hynix, and by extension in the health of the AI supply chain, are really paying for is the assumption that advanced memory stays scarce — protected by its technical complexity and by the American-led export-control regime that keeps China off leading-edge tooling. The espionage law is a lagging indicator of how much strain that assumption is now under.

The deeper implication is a change in what the memory business is. For four decades DRAM was a brutal cyclical commodity, and the standard fear was that when Chinese capacity finally arrived, prices would collapse and the cycle would resume as a race to the bottom. The espionage law, the Pentagon blacklist and Seoul's $880 billion plan to build a vast semiconductor hub all say the same thing: countries have decided the memory industry is a strategic asset to be defended with the apparatus of the state, not left to markets. Such a defence is a gift to the incumbents' shareholders while it holds. It is also the most honest available clue to how close the rents are to slipping — because nobody mobilises this much legal firepower over a moat they are confident will stand.