The Donald Trump administration has taken its next step in a trade offensive, this time targeting semiconductors.

Tariffs are the stated justification; the real objective is to force production onto U.S. soil. On Jan. 14, Trump signed a proclamation imposing a 25 percent duty on certain artificial intelligence (AI) chips. U.S. Commerce Secretary Howard Lutnick followed with a blunt warning: “Everyone who wants to build memory has two choices: They can pay a 100 percent tariff, or they can build in America.” Diplomacy has given way to coercion.

The target is unmistakable: memory semiconductors. DRAM and NAND, along with high-bandwidth memory (HBM) critical to AI, are squarely in the crosshairs. Samsung Electronics and SK hynix, which dominate the global market, are under direct pressure. For the first time, the U.S. has openly signaled its intent to pull memory production onto domestic soil, escalating the semiconductor battle to a new level.

The more serious problem is not the absence of assurances, but their emptiness. Last year’s Korea-U.S. joint fact sheet explicitly states that Korean chipmakers will receive “no less favorable” treatment than competitors — language intended to preserve most-favored-nation (MFN) status. Yet without defined quotas, benchmarks or enforcement mechanisms, those words offer little protection once leverage enters the equation.

That vulnerability was exposed by the U.S.-Taiwan semiconductor deal. By pledging roughly $250 billion in U.S. investment, Taiwanese firms secured near tariff-free quotas, with exemptions scaled to production capacity — 2.5 times during construction and 1.5 times after completion. The message is clear: MFN status alone is not enough. What matters is how much leverage a country brings to the table and how clearly that leverage is quantified.

For Korea, this sets a dangerous precedent. Samsung is already building a $37 billion wafer fab in Texas, while SK is investing $3.87 billion in an advanced HBM packaging facility in Indiana. At home, hundreds of trillions of won are being poured into the Yongin semiconductor cluster. Further demands for U.S. investment would not simply raise costs; they would risk a gradual hollowing-out of Korea’s industrial base.

Yet government responses remain largely rhetorical. Repeated references to the principle of MFN treatment are not backed by concrete numbers or benchmarks. In negotiations shaped by Trump-style tactics, abstract assurances are the riskiest strategy. As Trump wrote in his 1987 autobiography, “Don’t make deals without leverage. Enhance it.” Tariffs are his preferred instrument, and lip service is meaningless at the negotiating table.

Korea, however, is not powerless. Memory chips are the lifeblood of the U.S. AI industry. American firms cannot quickly replace Korean DRAM, NAND and HBM, and excessive tariffs would ultimately raise costs for U.S. tech companies and consumers alike. This supply-side reality is Korea’s leverage — if it chooses to use it.

What is needed now is decisiveness. MFN treatment must be fixed in enforceable terms, not merely cited as a principle. Tariff exemptions should be explicitly tied to investment levels, production capacity and supply-chain contributions, and benchmarked against the same standards applied to Taiwan. Safeguards are also needed to prevent past investments from being repackaged as leverage for new demands.

Memory chips account for roughly 30 percent of Korea’s exports and form the backbone of its manufacturing economy. If that base is destabilized, the broader economic consequences will follow. The Trump administration’s semiconductor tariffs are not a passing threat; they are the opening move in a long game. In such a contest, untested optimism is the most dangerous strategy of all. MFN status is not a matter of faith — it is a matter of execution.

Kang Seung-woo is the business desk editor at The Korea Times.