SK Hynix and Intel in Talks Over US Memory Chip Production

SK Hynix and Intel are reported to be in discussions about producing memory semiconductors within the United States. Neither company has officially confirmed the talks, and the precise terms and structure remain unclear. Yet if a deal is struck, analysts believe the consequences for global semiconductor supply chains could be far-reaching.

The backdrop: geopolitics and America's push for domestic chip production

The primary driver behind the negotiations is Washington's aggressive campaign to revive domestic semiconductor manufacturing. The CHIPS and Science Act, which came into force in 2022, offers tax credits of up to 25% on investments in American chip-making facilities, alongside billions of dollars in direct subsidies. The legislation has already drawn TSMC, Samsung, and Micron into building factories on US soil, and SK Hynix — South Korea's second-largest chipmaker — is under similar pressure to follow suit.

SK Hynix had already announced in 2023 a roughly $3.8bn investment in an advanced chip-packaging facility in Indiana. The reported discussions with Intel appear to extend that commitment further, moving beyond packaging to the actual manufacture of memory chips within the United States.

Intel's calculus: filling factories and finding customers

For Intel, the talks fit into a broader and increasingly urgent restructuring of its business. The American company has been attempting to expand its contract manufacturing operation, Intel Foundry Services (IFS), which produces chips designed by outside clients. Progress has been slow: TSMC and Samsung retain commanding positions in the foundry market, and Intel faces the pressing challenge of filling the large new factories it is building in Ohio and Arizona.

A deal in which SK Hynix's memory chips are manufactured in Intel's plants would offer both sides something valuable. Intel would gain the utilisation rates its new fabs desperately need; SK Hynix would satisfy American political demands for local production. Industry insiders suggest the arrangement could, at its most ambitious, extend to producing high-bandwidth memory (HBM) — the premium chips used in artificial-intelligence hardware — on American soil.

AI demand: a further spur to agreement

The explosive growth of the AI accelerator market adds another dimension to the talks. As Nvidia, AMD, and other chip designers scramble to secure supplies of HBM, SK Hynix has emerged as the dominant global supplier. According to TrendForce, a market research firm, SK Hynix accounted for more than 50% of the global HBM market in 2024.

For American AI companies, building a secure domestic supply chain for HBM has become a strategic priority. Both the Biden and Trump administrations have emphasised the importance of onshoring critical components for AI, giving SK Hynix's potential move into US memory production strong political tailwinds.

Obstacles: costs, technology transfer, and competition

Whether the negotiations will reach a successful conclusion is far from certain. Profitability is the most immediate concern. Labour and operating costs in the United States are substantially higher than in South Korea or Taiwan, and even generous subsidies may not fully offset the resulting increase in unit production costs. Micron, which is already expanding its American manufacturing footprint with $6.1bn in federal subsidies for facilities in Idaho and New York, has repeatedly flagged the financial burden of domestic production.

Technology transfer poses a further complication. Manufacturing SK Hynix's DRAM and HBM chips in Intel's facilities would require a significant degree of process know-how to be shared, raising delicate questions about intellectual-property protection that could complicate and potentially derail the negotiations.

The competitive landscape is also shifting. Samsung is constructing a foundry plant in Taylor, Texas. Should SK Hynix establish its own US memory production capability, competition in the American market would intensify considerably.

Lessons from Japan and Europe

Precedents for foreign chipmakers producing semiconductors in partnership with local firms already exist elsewhere. TSMC built its "JASM" factory in Kumamoto, Japan, in a joint venture with Sony and Denso; the plant began full operations in 2024. In Europe, TSMC adopted a consortium model with Infineon and NXP for its Dresden facility in Germany. Such structures distribute the capital burden while generating goodwill with host governments.

A similar arrangement — a joint venture or contract-manufacturing agreement in which technology and capital are shared — seems the most likely outcome of the SK Hynix-Intel talks. That said, memory chips present particular difficulties that logic chips do not. Process standardisation is paramount in memory manufacturing, and cost competition is exceptionally fierce, making the design of a viable profit model the central challenge in any negotiations.

Outlook

With the talks still at an early stage and both parties declining to confirm them, sweeping conclusions would be premature. The possibility of a breakdown remains real. Even so, the very fact that such negotiations are reportedly under way carries a clear signal: the geographic diversification of memory chip production has moved from a strategic option to a near-inevitability.

Experts argue that, regardless of how the current talks conclude, SK Hynix's expansion of its American manufacturing presence is a matter of when, not if. Three structural forces — US industrial policy, surging AI demand, and geopolitical risk — are pushing in the same direction simultaneously.

For South Korea's semiconductor industry, the central challenge is how to satisfy American demands for local production without surrendering the technological edge that makes its chipmakers indispensable in the first place. Whatever the outcome, these negotiations will stand as a significant marker in the ongoing reorganisation of the global memory chip supply chain.