The geography of the memory-chip industry has been set for decades, with the bulk of production concentrated in South Korea. That map may be about to shift, at least by one factory. Solidigm, the American subsidiary of SK Hynix, is evaluating whether to build a NAND flash plant in the United States, with upstate New York as the leading candidate, according to a Reuters report on September 18 that cited three people familiar with the matter.
The project would be separate from another set of talks the company is holding, those between SK Hynix and Intel over producing memory at Intel’s Ohio site. The two ideas run in parallel, and both point the same direction: the South Korean memory giant is looking for ways to put production on American soil without exposing its most sensitive technology to political risk.
The choice of NAND is the key to the strategy, the sources said. NAND flash is the memory used for storage, and it is regarded as less sensitive than high-bandwidth memory or other advanced DRAM, the chips that sit at the leading edge and draw the attention of export-control regulators. Putting a NAND plant under an American subsidiary carries less political resistance than moving the more advanced technology across the water.
Solidigm is the vehicle for that logic. The company is the flash-memory business that SK Hynix assembled from Intel’s NAND operations, an American name with American management and an American footprint. Routing a U.S. fab through Solidigm gives SK Hynix a domestic face for the project, which matters when the factory will sit in a state whose senators care about which companies are building there.
Upstate New York has spent years courting chipmakers, and the region already carries the marks of that effort. The state has offered large incentive packages to semiconductor projects, and the memory industry has taken notice. A NAND plant would add a different kind of capacity to a region that has so far drawn more logic and foundry investment than memory.
There is no final agreement, and the sources were careful to say so. The talks are early, the location is a preference rather than a decision, and the economics are unresolved. Cost is the main obstacle. Building a fab in the United States runs clearly higher than building one in Korea, and memory is a business where cost per wafer decides who survives a downturn.
That cost gap is the reason the project is still a question rather than an announcement. The U.S. government has offered subsidies to close the gap, but subsidies are not enough to erase the difference in labor, construction and energy costs that separate an American site from a Korean one. A company has to believe the political and strategic benefits are worth the premium, and that calculation has not been finished.
The political benefit is real. The United States has pushed hard to bring advanced chip manufacturing back within its borders, and memory has been the missing piece of that push. Logic and foundry capacity has drawn the biggest subsidies, while memory, which is capital-intensive and price-cyclical, has been harder to place. A NAND plant would fill part of that gap.
The strategic benefit is real as well. The memory industry has always been exposed to the geopolitics of the region where it is concentrated, and every major producer has been weighing whether to hedge with capacity elsewhere. An American fab is a hedge, a way to keep selling into the American market even if the politics around the home country shifts.
Solidigm’s own position makes the project natural. The company already has operations and employees in the United States, and it already sells into the American market under its own name. Adding a factory is an extension of an existing American business rather than the creation of a new one, and that continuity lowers the political and operational friction.
NAND is also a business in transition. The market for flash memory is deeply cyclical, swinging between shortage and glut on a schedule that no company controls, and the survivors tend to be the ones with the lowest costs. An American plant, with its higher costs, would need either subsidies or a strategic buyer to justify itself through the down cycles. That is the arithmetic the company is still working.
The decision will come down to that arithmetic. If the subsidies and the strategic value outweigh the cost premium, the plant gets built. If the cost premium wins, the map stays where it is, and the memory industry keeps its concentration in Korea. The sources said no final agreement exists, and for now the calculation is still being done.
What the report does establish is that the conversation is serious. Three people familiar with the matter would not be describing a leading site and a political strategy for a project that exists only on paper. The question is not whether SK Hynix wants an American plant, but whether it can make the numbers work well enough to announce one. For the memory industry, and for upstate New York, that question is now on the table.
