Solidigm eyes US factory for NAND chips, upstate New York considered

Solidigm, SK Hynix's subsidiary, is reviewing construction of a U.S. NAND flash manufacturing plant, with upstate New York under consideration, though no agreement or final project details have been confirmed. The potential investment would diversify production beyond Solidigm's sole NAND site in Dalian, China, while reducing exposure to potential U.S. tariffs and export restrictions on China. AI data-center investment has intensified the global memory-chip shortage and raised the urgency for new capacity, but substantially higher U.S. manufacturing costs and ongoing U.S.-South Korea trade negotiations remain key constraints.

Analysis

The market is likely to over-attribute any Solidigm U.S. capacity decision to INTC. Unless Intel discloses a binding wafer-services agreement, committed volume, pricing, and customer-funded capex, the economic contribution to Intel Foundry is immaterial relative to its execution and utilization gap; headline-driven INTC strength should be faded rather than treated as validation of the Ohio ramp.

For SK Hynix (000660 KS), geographic redundancy has strategic value but is unlikely to improve consolidated margins near term. A greenfield U.S. NAND fab would carry materially worse labor, construction, and utilities economics than Korean capacity, making subsidies, tariff avoidance, and long-duration customer commitments prerequisites—not upside. The first-order beneficiary is less NAND supply than equipment and construction demand, but revenues for AMAT, LRCX, KLAC, and local infrastructure suppliers would not begin to matter until a final investment decision and tool-order cycle, likely 12-24 months away.

The contrarian point is that AI-driven memory tightness is predominantly HBM/DRAM-led; NAND pricing can improve through enterprise SSD demand, but a prospective new fab does not resolve the binding AI-memory constraint. Over 6-18 months, incremental U.S. NAND capacity would be a competitive negative for pure NAND exposure such as SNDK if industry utilization weakens, while MU retains the stronger policy and domestic-memory optionality. Thesis falsifiers are a binding capacity agreement with Intel, disclosed U.S. subsidies sufficient to offset cost disadvantage, or NAND contract pricing sustaining increases despite announced incremental capacity.

Key Decisions for Investors

• Do not add directional INTC exposure on this report. Treat a >5% headline-driven move without disclosed contract economics, wafer volumes, or prepayments as a tactical fade; cover if Intel announces a funded multi-year agreement or raises foundry utilization guidance.
• Maintain/establish a 6-12 month long MU / short SNDK relative-value position only if NAND spot and contract pricing continue to rise while DRAM/HBM pricing remains firmer. The setup captures MU's superior domestic-policy optionality and lower pure-play NAND supply risk; exit if NAND utilization tightens materially or SNDK secures customer-backed U.S. capacity.
• Place AMAT, LRCX, and KLAC on a 12-24 month capex alert rather than buy now. Initiate only following a final investment decision plus identified tool vendors and subsidy funding; a credible U.S. NAND fab could create a multi-year equipment order tail, but current information has no revenue timing or capex envelope.
• For SK Hynix (000660 KS), retain a core AI-memory exposure but avoid assigning value to a U.S. NAND plant until management quantifies subsidy support and return thresholds. A project proceeding without material grants or customer prepayments would be a margin/ROIC negative despite lower geopolitical concentration.