For enterprise storage buyers, the distinction is important. A U.S. fab could eventually give Solidigm a second manufacturing geography beyond Dalian, China, and reduce exposure to tariffs or equipment-export restrictions. It will not add NAND supply, alter SSD availability, or change procurement plans in 2026 or 2027. Reuters notes that a new semiconductor plant normally requires years before production begins, even after a final investment decision.
The report is also separate from SK hynix’s recently disclosed talks with Intel over possible U.S. memory manufacturing at Intel’s Ohio facilities. Those discussions concern a different potential arrangement; they are not a shortcut to putting Solidigm NAND into production in New York.
What Solidigm and SK hynix have actually confirmed
Neither company has announced a U.S. NAND fab. In its response to Reuters, SK hynix said Solidigm is reviewing various ways to improve its competitiveness but that no specific plans are confirmed. Solidigm said it does not comment on market rumors and added that no decision on a particular course of action has been made.
That language is substantially more cautious than the headline-level claim that Solidigm “plans” a plant. The current state of play is a strategic review, not a capital commitment. There is no announced project cost, technology node, clean-room capacity, equipment supplier, government incentive package, customer commitment, or targeted start of construction.
Seoul Economic Daily independently reported on September 18 that Solidigm was reviewing candidate sites and terms for a U.S. East Coast NAND mass-production line. Reuters was more specific in calling upstate New York a leading contender. Those accounts are compatible—upstate New York is an East Coast location—but neither establishes that New York has been chosen.
The absence of a firm commitment is especially relevant because NAND fabs are unusually capital-intensive bets in a market that has repeatedly swung between shortage and oversupply. A proposal based on today’s AI-driven storage demand would have to remain viable when a facility begins shipping wafers several years from now.
This would address a real Dalian concentration risk
A U.S. fab would matter because Solidigm’s NAND supply remains centered on the former Intel facility in Dalian, China. Intel announced the sale of its NAND and SSD business to SK hynix in October 2020; the first closing occurred on December 29, 2021, transferring Intel’s SSD business and the Dalian NAND manufacturing facility. The remaining wafer-manufacturing IP and personnel transferred in the final closing in 2025.
That timeline corrects a simplification in the submitted reporting: SK hynix did not simply acquire an already complete Solidigm manufacturing operation in 2021. Solidigm was created as the U.S.-based SSD business following the first phase of the Intel deal, while the Dalian fab itself moved to SK hynix. Blocks & Files has reported that the Chinese fab is operated by an SK hynix Chinese subsidiary rather than by Solidigm directly.
The organizational distinction may sound technical, but it explains why a prospective U.S. plant is more than a new Solidigm address. It would represent a move toward a geographically separate wafer-production base for the NAND products that feed Solidigm’s enterprise SSD portfolio.
SK hynix has already expanded its U.S. presence in areas other than NAND wafer fabrication. The company is building an advanced packaging facility in West Lafayette, Indiana, and announced in January that it would restructure Solidigm into a broader U.S.-based AI-focused business initiative, provisionally called AI Company. The proposed NAND fab would add front-end memory production to that footprint—but it remains only a possibility.
Dalian expansion makes the timing less straightforward
The potential U.S. move comes while SK hynix is still adding capacity in China. Reporting in August from Blocks & Files, citing South Korean coverage, said work had resumed on Dalian Plant 2 after an earlier pause, with wafer production expected to begin in the first half of 2027 and eventually reach about 50,000 wafers per month.
If that schedule holds, it provides the nearer-term capacity increase. The same report said Dalian Plant 1 produces roughly 100,000 wafers per month, making Plant 2 a meaningful, though not transformational, addition. It also reported that the new Dalian line would use mature floating-gate NAND technology in the roughly 100-layer class, rather than SK hynix’s more advanced charge-trap NAND produced in South Korea.
That undercuts one implication sometimes attached to the U.S.-fab report: a domestic facility would not necessarily be built around Solidigm’s most advanced or highest-density NAND. No one has disclosed the intended process technology. NAND manufacturing equipment, yield expertise, IP arrangements, qualification cycles, and customer needs would all determine what such a plant could make.
For IT buyers, the technology question matters more than the factory’s postal address. Solidigm’s appeal in the data center has increasingly centered on high-capacity QLC enterprise SSDs, where cost per terabyte and rack density can outweigh the endurance and latency advantages of TLC NAND. A U.S. fab would not automatically produce the same NAND generation used by its current flagship drives, nor would it guarantee domestic sourcing for a particular SSD model.
AI demand explains the interest, not the business case
Reuters and Seoul Economic Daily both point to AI data centers as a key reason for revisiting NAND expansion. That is credible as a demand signal: training, inference, retrieval systems, checkpoints, logs, and data-preparation pipelines all increase storage consumption, while high-capacity QLC SSDs reduce the number of drives, servers, slots, and watts needed to deploy a given amount of flash.
But “AI demand” alone is not a factory plan. NAND pricing has historically been volatile, and a large fab cannot be rapidly slowed or redirected once built. Reuters reported that SK hynix itself acknowledged the uncertainty around long-term demand by the time a new plant could come online.
The U.S. policy environment is another evident factor. Reuters reported that a domestic plant could reduce Solidigm’s dependence on Dalian while limiting exposure to potential U.S. tariffs and restrictions affecting the export of chipmaking equipment to China. The company is therefore weighing a resilience calculation as much as a capacity calculation.
There is also a political difference between NAND and SK hynix’s highest-value DRAM products. Reuters reported that its sources viewed NAND production as less politically sensitive for South Korea than advanced DRAM or HBM production. That could make a Solidigm NAND project easier to discuss than an American HBM fab, but it does not settle the economics, technology transfer, or incentives needed to build one.
What storage teams should watch for
The immediate operational advice is simple: treat this as a strategic signal, not a supply commitment. Organizations planning enterprise SSD refreshes should continue to qualify drives based on endurance ratings, firmware support, controller behavior, warranty coverage, availability allocations, and total cost per usable terabyte—not an uncommitted U.S. manufacturing proposal.
The next meaningful evidence would be concrete and difficult to mistake for speculation:
- SK hynix or Solidigm would need to announce a final investment decision, location, project cost, process technology, and production target.
- New York State or federal agencies would likely disclose incentives, permitting, environmental review, or infrastructure agreements if a specific site advances.
- Large cloud or server customers could emerge as anchor buyers if the project is designed around AI data-center storage demand.
- Solidigm would need to explain whether the fab would supply existing enterprise SSD lines, a future QLC generation, or a new product family.
Until those details appear, the important development is not that Solidigm is building an American fab. It is that SK hynix is openly examining a way to reduce the NAND business’s dependence on Dalian while AI infrastructure spending and U.S.-China technology restrictions make that dependence harder to ignore.