SK Hynix’s talks with Intel over using an Ohio manufacturing facility could give the United States a faster route to expanding AI-memory production, while testing whether advanced South Korean semiconductor technology can be manufactured abroad. The discussions reportedly include leasing part of Intel’s Ohio factory, forming a joint venture with possible participation from major cloud companies, or adopting another structure that would allow SK Hynix to use Intel’s US manufacturing infrastructure.
The proposal is not final, and the companies are still considering its business, manufacturing and technology terms. But the talks reveal a larger shift in the geography of the AI supply chain. Demand for high-bandwidth memory, or HBM, has risen alongside the expansion of AI data centres, putting pressure on memory manufacturers to increase capacity while governments seek greater domestic control over strategically important chips.
For Intel, the discussions are linked to the company’s effort to attract major external customers to its foundry business. Intel has been trying to turn contract manufacturing into a central part of its turnaround strategy under Chief Executive Lip-Bu Tan. A deal with SK Hynix would provide a large customer for its US manufacturing operations and potentially improve the utilisation of facilities that Intel is seeking to position as part of a broader domestic chipmaking base.
For SK Hynix, the Ohio option would offer a different path from constructing an entirely new plant. Leasing space inside an existing Intel facility could allow the South Korean company to expand its US manufacturing footprint without relying only on a new site. The arrangement could also give Intel a stronger role in the production network supporting AI processors, although the precise manufacturing model remains under discussion.
### Why the SK Hynix Ohio factory proposal matters
HBM is a specialised form of memory designed to move large amounts of data at high speed. It is a key component of Nvidia’s AI chips, according to the report, because AI processors need to handle substantial volumes of data during computing operations. As cloud companies build more data centres and companies increase their use of AI systems, demand for the memory supporting those processors has grown sharply.
That demand has already influenced SK Hynix’s expansion plans. The company announced in June that it intended to double its production capacity over the next five years as it prepared for continued demand from AI and data centres. Chairman Chey Tae-won said in July that demand for HBM was “enormous”, underlining the commercial pressure behind the company’s search for additional production capacity.
The company has also begun developing its own US manufacturing presence. SK Hynix broke ground last month on its first US facility in Indiana. Chief Executive Kwak Noh-Jung said the plant was expected to make Indiana a key HBM production base in the United States by 2030. The Indiana project represents a conventional expansion model: the company is developing a dedicated new facility for its operations.
The Ohio discussions represent a more flexible model. Rather than waiting for a new factory to be completed, SK Hynix could potentially use an existing Intel site. The distinction matters because semiconductor capacity is not created only through corporate investment announcements. It depends on access to suitable facilities, specialised equipment, process technology, trained workers, reliable utilities and customers able to absorb the output. Using existing infrastructure could address part of that challenge, although the reported talks do not establish whether Intel’s Ohio facility can immediately support the specific requirements of advanced memory production.
### A new role for Intel’s US manufacturing network
The proposed arrangement would also be important for Intel’s foundry ambitions. A foundry manufactures chips for outside customers rather than only producing a company’s own designs. Intel has been seeking large customers as it builds this business, and SK Hynix would represent a significant addition to its external manufacturing base if an agreement were reached.
A lease, joint venture or another partnership would distribute responsibilities differently. In a lease model, SK Hynix could occupy part of Intel’s facility and operate its own production activity. A joint venture could place the companies in a shared structure, potentially with cloud companies involved as customers or partners. Other arrangements could give Intel a different role in manufacturing, infrastructure or services. The source report does not establish which model the companies will choose.
This uncertainty is central to understanding the proposal. Semiconductor manufacturing partnerships are not ordinary industrial leases. They require agreement on ownership, process control, equipment, intellectual property, workforce responsibilities, output allocation and technology safeguards. They must also fit the technical characteristics of the products being manufactured. The Ohio plan therefore remains a framework under discussion rather than a confirmed production project.
For Intel, the talks suggest that the value of its manufacturing assets may depend partly on whether they can host production for companies with different technology requirements. For SK Hynix, the talks show how access to existing manufacturing infrastructure could complement, rather than replace, its own long-term capacity expansion. The two companies have different strategic needs, but the proposed partnership could connect them through the demand created by AI computing.
### The South Korean technology constraint
The main reported obstacle is not only commercial. Advanced memory technologies, including technology related to HBM, are considered sensitive in South Korea. Reuters reported, citing three sources, that the South Korean government could oppose an arrangement involving production of advanced memory such as HBM in the United States.
That issue places the possible Ohio project within a wider tension in semiconductor policy. The United States is pushing to expand domestic chip production, while South Korea has an interest in retaining control over strategically important technologies developed by its companies. SK Hynix’s decision about what can be manufactured in the United States would therefore have implications beyond the company’s property and capacity choices.
The reported government concern also helps explain why the talks cannot be treated as a straightforward response to rising demand. Even if Intel has available space and SK Hynix needs additional capacity, the companies would still need to determine which products and technologies could be transferred, where critical production steps would occur and what safeguards would apply. The source material does not say whether South Korean authorities have issued a formal decision or what conditions they might attach to any approval.
This is an institutional question as much as a corporate one. The future of AI hardware is being shaped simultaneously by company expansion plans, national industrial strategies and controls over sensitive technologies. Manufacturing location is no longer only a matter of operating cost or proximity to customers. It is also linked to technology governance and supply-chain security.
### AI demand is changing semiconductor infrastructure
The immediate commercial driver is the rapid expansion of AI data centres. These facilities require large volumes of processing power and memory, and the resulting demand has contributed to a shortage of advanced memory chips, according to the report. SK Hynix’s plan to increase capacity over five years and its US projects are responses to that demand environment.
The infrastructure implications extend beyond the factory itself. Semiconductor plants depend on large, specialised industrial sites and a stable ecosystem of equipment suppliers, technical workers, logistics networks and utility services. The decision to use an existing facility can reduce the need to develop an entirely new site, but it does not remove the need to adapt production systems and secure the supporting infrastructure required for a different type of chip.
The comparison between Indiana and Ohio highlights two models of industrial expansion. Indiana is being developed as SK Hynix’s own facility and is expected to become a US HBM base by 2030. Ohio could provide access to Intel’s existing manufacturing platform through a lease, joint venture or another arrangement. One model prioritises dedicated control over a new site; the other explores shared or repurposed capacity.
Neither model is automatically faster or simpler. A new facility requires substantial construction and commissioning, while an existing site may require technical modifications, equipment changes and agreement over operating responsibilities. The reported discussions do not provide a timeline for the Ohio proposal, so it is not possible to establish whether it would begin production before or alongside the Indiana facility.
### The bigger industrial-policy question
The proposed SK Hynix-Intel arrangement shows how the AI boom is altering the relationship between technology companies and physical infrastructure. AI expansion is often discussed through software, computing models and data, but the system also depends on factories capable of producing specialised processors and memory at scale. The location and ownership of those factories are becoming strategic questions for both companies and governments.
The talks also illustrate why domestic manufacturing targets cannot be measured only by the number of factories announced. Capacity must be matched with the right process technology, customers, equipment, workforce and regulatory permissions. A facility can become strategically useful only when those elements are aligned.
For now, the evidence confirms that SK Hynix and Intel are examining ways to use Intel’s US manufacturing facilities, including a possible Ohio lease or joint venture. It also confirms that the proposal faces a potential technology-transfer obstacle in South Korea and that SK Hynix is pursuing a separate Indiana expansion. What remains unresolved is the structure of any deal, the products that could be made in Ohio, the role of cloud companies and whether South Korean authorities would permit advanced memory production there. Those decisions will determine whether the proposal becomes a working part of the US AI supply chain or remains an exploratory partnership.
