Exclusive-SK Hynix in talks with Intel about deal to make memory chips in the US for the first time, sources say

SK Hynix is in exploratory talks with Intel to manufacture memory chips in the U.S., potentially through leasing space at Intel's delayed Ohio fab or forming a venture with Intel and cloud companies. The potential deal could provide Intel financial relief and expand U.S. supply of AI-critical HBM and other memory chips, but advanced-chip production may face South Korean national-technology reviews. Intel's Ohio project was initially planned as up to a $100 billion investment, although its two plants have been delayed until 2030 and 2031; U.S. tariff threats of up to 100% add pressure for Korean and Taiwanese chipmakers to localize production.

Analysis

The equity implication for INTC is less about incremental foundry revenue than monetizing stranded Ohio infrastructure and reducing fixed-cost absorption risk. A tenant/JV structure could improve the site’s capital efficiency, but memory and leading-edge logic require materially different process tools and supply ecosystems; unless the arrangement is primarily shell, utilities, and shared services, conversion capex could dilute the apparent benefit. Market should therefore discount any headline until lease economics, customer volume commitments, and the party funding incremental tools are disclosed.

SK Hynix’s strategic upside is customer proximity and tariff-risk mitigation, particularly if hyperscalers provide take-or-pay support. The offset is structural: a U.S. memory cost base risks lower gross-margin capture than Korean output, while Korean technology-transfer approval could limit the products to mature DRAM or packaging rather than high-value HBM. That distinction matters for MU: a U.S.-based SK Hynix HBM line would increase competitive intensity in the most profitable AI-memory pool over 12-24 months; a mature-memory facility would be largely immaterial to current HBM pricing.

Near term, this is an INTC sentiment catalyst rather than an earnings catalyst, and the press-report premium is vulnerable to a non-binding outcome. Over 1-3 months, the key read-through is whether cloud customers commit capital or long-duration purchase agreements; their participation would validate localized memory supply as an AI infrastructure bottleneck rather than a political concession. Contrarian view: Washington’s leverage may produce an investment announcement, but Seoul’s review process and cost burden make a smaller packaging/assembly or phased project more probable than a fully capable HBM fab.

Key Decisions for Investors

• Treat any INTC rally on an announcement as tactical unless disclosed consideration materially offsets Ohio carrying costs: buy only after terms identify committed tenant payments/customer prepayments; exit if no definitive agreement or funding framework emerges within 90 days.
• Maintain MU as the cleaner U.S. memory scarcity exposure over INTC for the next 1-3 quarters. Reassess if SK Hynix confirms U.S. HBM wafer production with hyperscaler offtake, which would be a 12-24 month margin and valuation-multiple headwind for MU.
• For event-driven exposure, use a defined-risk INTC call spread dated 3-6 months rather than outright stock, sized to the probability of a definitive agreement. Falsifier: disclosures showing only exploratory discussions, no third-party capital, or a project limited to non-fab operations.
• Avoid treating NDAQ as a material beneficiary: a secondary listing does not change SK Hynix’s manufacturing economics or create meaningful exchange revenue sensitivity absent sustained U.S. liquidity and institutional trading volume.