SK hynix is weighing a move that could reshape where one of the world’s most important memory components is manufactured. Its U.S.-based Solidigm subsidiary is considering building its first NAND flash memory factory in the United States, with upstate New York emerging as a leading candidate, according to three people familiar with the discussions cited by Reuters.
The proposal is still preliminary. SK hynix said Solidigm is reviewing options to strengthen its competitiveness but that no specific plan has been confirmed. Solidigm separately said it does not comment on market rumours or speculation. That distinction is important: this is not yet a committed U.S. fab project, but it signals how quickly AI-driven memory demand and semiconductor geopolitics are changing investment decisions.
Why Solidigm Is Looking at the U.S.
Solidigm is the NAND and SSD business created after SK hynix acquired Intel’s NAND and solid-state drive operations. When the first phase of that transaction closed, SK hynix acquired Intel’s Dalian NAND flash manufacturing facility in China, while Solidigm became the U.S.-based brand managing the acquired storage business. SK hynix’s own announcement described the Dalian facility as part of the acquired NAND manufacturing assets.
A new U.S. plant would therefore change the geographic balance of Solidigm’s manufacturing footprint. It could reduce the company’s dependence on Dalian while making its supply chain less exposed to U.S.-China trade restrictions, tariffs and controls on semiconductor manufacturing equipment.
The timing is significant. Washington is pushing foreign semiconductor companies to manufacture more inside the United States, while the global memory market is being pulled in the opposite direction by enormous demand from AI infrastructure. Data centres increasingly need both high-bandwidth memory for accelerators and large amounts of conventional DRAM and NAND storage for training, inference, datasets and system storage.
AI Is Turning NAND Into Strategic Infrastructure
NAND flash is the non-volatile memory used in SSDs and many other storage products. It is different from the high-bandwidth memory attached directly to AI accelerators, but the two markets are becoming increasingly connected as AI systems process larger datasets.
Solidigm itself has been positioning enterprise SSDs as part of AI infrastructure. In 2025, the company said its high-capacity SSD portfolio was being used for AI and data-intensive workloads, including a 122TB data-centre SSD. More recently, Solidigm said it had expanded its Sacramento-area R&D operations and NAND laboratory to support AI data-centre development. The company’s April 2026 announcement said it had added nearly 100 NAND tools to its R&D environment.
That matters because AI servers do not simply consume GPUs. They require a hierarchy of memory and storage capable of feeding those processors with data. As model sizes, datasets and inference workloads expand, high-capacity SSDs can become an important part of the overall system architecture.
China Dependence Is Becoming Harder to Ignore
Solidigm’s existing Dalian manufacturing base sits at the centre of another strategic problem: access to advanced semiconductor technology is increasingly shaped by U.S. export controls.
Reuters reported that a U.S. factory could reduce Solidigm’s reliance on its Dalian facility while limiting exposure to potential restrictions on equipment exports to China. The company is also facing intensifying competition from Chinese memory manufacturers. ChangXin Memory Technologies, or CXMT, is preparing to enter NAND production as it seeks to compete with established players including Samsung, SK hynix and Micron.
That competition is arriving while memory supply is tightening. AI data-centre construction has increased demand for memory across the industry, making manufacturing location a strategic question rather than simply a cost calculation.
SK Hynix Is Expanding NAND Capacity Elsewhere Too
The potential Solidigm fab should not be viewed as SK hynix’s only NAND expansion plan. In August, SK hynix announced a roughly 54 trillion won investment across new facilities in Yongin and Cheongju in South Korea. Its Cheongju M17 project is specifically intended to expand NAND production, with the first cleanroom targeted for December 2028.
SK hynix said NAND demand is being driven increasingly by enterprise SSDs as AI services expand. It also highlighted the growing role of storage in AI inference, including workloads involving large key-value caches.
The contrast is revealing. South Korea remains a core manufacturing base where SK hynix can expand established operations, while a possible Solidigm fab in the U.S. would serve a different strategic purpose: geographic diversification and closer alignment with American customers and policymakers.
What It Could Mean for India
India is not directly involved in the proposed project, but Indian consumers and electronics manufacturers are exposed to the same global memory cycle.
India imported $6.79 billion worth of memory chips in FY26, according to Business Standard data, with the import bill rising sharply as memory prices increased. CyberMedia Research has also reported that India’s consumer flash-storage shipments fell 27% year over year in the first quarter of 2026 as higher NAND prices affected the market.
That makes global memory capacity relevant to everything from smartphones and laptops to SSDs and data-centre infrastructure in India. A U.S. NAND plant would not suddenly lower Indian storage prices, especially because no project has been approved yet. But greater geographic diversification could eventually make the global supply chain more resilient when geopolitical disruptions affect a single manufacturing region.
India is simultaneously trying to build more of its own semiconductor supply chain. At SEMICON India 2026, the country highlighted new investments spanning fabs, packaging, equipment and materials. That strategy is aimed not only at producing chips domestically, but at reducing the vulnerability created when critical components must travel through concentrated global supply chains.
The Bigger Story Is Supply-Chain Geography
The most important point about the Solidigm proposal is not the location of a possible New York factory. It is what the discussion says about the semiconductor industry’s next phase.
For years, chip manufacturing was optimised around cost, technical expertise and established industrial clusters. Those factors still matter, but governments and customers now increasingly value resilience, domestic capacity and geopolitical security. AI is adding another pressure: companies need more memory than many supply chains were designed to provide.
If Solidigm ultimately builds a U.S. NAND fab, it would become another example of memory production being distributed across more regions. If it does not, the fact that SK hynix is seriously evaluating the option still shows how strategic NAND manufacturing has become.
No Factory Has Been Approved Yet
For now, the project remains under consideration. Reuters’ sources said upstate New York is emerging as a leading candidate, but key details such as investment size, construction timing and technology scope have not been finalised.
That uncertainty should not obscure the larger signal. AI is driving demand far beyond GPUs, and memory companies are being pushed to rethink where they manufacture the storage layer that supports modern computing. Solidigm’s possible first U.S. NAND fab is therefore less a finished project than an early indicator of where the semiconductor supply chain may be heading next.
