SK hynix is seeking permission to export additional US-made semiconductor manufacturing equipment to its Dalian plant in China as a condition for building a NAND flash plant in the United States, Reuters reported on October 15. The company wants to increase its supply of SSDs for AI data centers, while the US government is simultaneously trying to attract factories to the US and curb expansion of production capacity in China. SK hynix, however, responded that equipment installation in Dalian is proceeding with US government approval and that it is not seeking to expand production capacity. The report points to a plan in which securing supply and revenue at existing sites in China would underpin future US production, but also to the export-control hurdles that stand in the way first.
Is Equipment for Dalian About "Maintaining Operations" or "Expanding Output"?
According to the negotiations as described by two people familiar with the matter, Reuters reports that US-based Solidigm, an SK hynix subsidiary, wants to install equipment in an unused section of the Dalian plant to increase NAND production. NAND is the memory element that retains data even when power is off, and it provides the storage capacity of SSDs. The idea of a US plant had already been reported in September, but what is new this time is that it links permission to export equipment to China with investment in the US.
What complicates the negotiations is a policy dated August 29, 2025 from the US Department of Commerce's Bureau of Industry and Security (BIS). BIS revoked the Validated End-User (VEU) exemption, under which some foreign companies could receive US-origin equipment and technology at their own plants in China without individual export licenses.
At that time, BIS indicated it intends to grant licenses needed to keep existing plants operating, while explicitly stating that it will not approve licenses for capacity expansion or technology upgrades. The measure does not block all equipment shipments to Chinese plants. The intended use of the equipment becomes the key point in deciding whether to grant a license.
The explanation around Dalian's "Mod B" overlaps with this dividing line. SK hynix told Reuters that the section was completed in 2023 and that equipment installation is proceeding in close consultation with, and with approval from, the US government. It added that it is not seeking to expand Mod B's production capacity. There is a gap between the expansion aim described by the sources and the position the company has stated.
That said, the fact that the building has existed for some time does not by itself settle whether additional equipment would count as a capacity expansion under export controls. Without knowing the criteria for the equipment and production capacity a license would cover, the gap cannot be closed. According to Reuters, the Commerce Department did not respond to its inquiry, and there has been no announcement that the US government has approved this request.
Three Investment Plans Differ in Process and Degree of Certainty
SK hynix is already making separate investments in the US. The Indiana plan announced in April 2024 is a $3.87 billion advanced packaging and R&D facility for HBM. HBM stacks DRAM chips to secure large bandwidth, supporting high-speed data processing close to GPUs.
The Indiana investment covers advanced HBM packaging and Cheongju M17 is a NAND manufacturing plan, so both differ in process and certainty from the proposed new US NAND plant.
Arranged by target process and decision status at the time of each announcement, based on disclosures confirmed as of October 16, 2026, the plans look like this:
| Plan | Production function added | Disclosed details and degree of certainty |
|---|---|---|
| Facility in Indiana, US | Advanced HBM packaging and R&D | A $3.87 billion investment plan was announced in an SK hynix release on April 3, 2024. It is not a NAND manufacturing plant |
| M17 in Cheongju, South Korea | NAND manufacturing base | A 19.1 trillion won investment was decided in an SK hynix release on August 7, 2026. Construction is targeted to start in February 2027, with the first cleanroom opening in December 2028 |
| Solidigm's proposed US NAND plant | NAND manufacturing in the US | A Reuters report on October 15 described its link to permission for equipment going to China. It has not been announced as a plan with a confirmed location, investment amount, or mass-production timing |
The table does not compare production volumes or how soon operations would begin. In particular, M17's date is a target for opening the cleanroom, not the start of NAND mass production. On memory manufacturing in the US, the company stated in a September 16 statement that specific plans, including cooperation with Intel, had not been decided. That statement reflects the decision status at the time and does not deny the negotiations reported now.
The difference in process matters when considering the effect on supply. At a NAND plant, memory elements are built on wafers. Advanced HBM packaging is the process of stacking and connecting manufactured DRAM chips to finish the product. Even if HBM back-end operations are set up in the US, that alone does not increase the NAND used in SSDs. Simply adding up US investment amounts does not reveal which component's supply capacity will grow.
AI SSDs Need Supply Before a New Plant Is Ready
On August 5, CoreWeave signed a multi-year strategic agreement with Solidigm, securing priority access to enterprise SSD supply capacity. CoreWeave itself explains that storage has become a constraint on its capacity planning and must be secured in line with rising AI demand. Customers are starting to lock in supply allocations for when they need them rather than simply waiting for future plants to be completed.
In announcing the M17 investment, SK hynix also explained that, beyond the growth of AI services, demand to store KV cache during inference is pushing up demand for enterprise SSDs. KV cache is a mechanism that stores the Key and Value information the AI has computed, reducing the burden of repeating the same calculations. As the amount of data to store grows, how to allocate data between high-speed memory and large-capacity storage also becomes an issue. This does not mean SSDs run at the same speed as HBM; it means their role in retaining necessary data is expanding.
On the other hand, a new NAND plant cannot increase supply immediately once construction is decided. Even for M17, the target for opening the first cleanroom is December 2028. SK hynix says it will build the plant according to plan while expanding cleanrooms and installing equipment in stages in line with demand. There are further steps between completing a building and actually increasing manufacturing capacity.
One of Reuters' sources explained that increased production in China would help secure funding for the US plant and serve as a bridge until output can increase in the US. The idea is to use existing sites to meet current demand and channel the profits to future sites. However, how much additional equipment would raise output and how funds would be directed to the US plant have not been disclosed. It is not yet possible to quantify how much faster or cheaper using Dalian could expand supply compared with a new plant.
The Question Is Where to Allocate Investment, Not a Funding Shortfall
In a statement on Solidigm dated October 1, SK hynix explained that it has a strong financial base and that using internal funds is an option. It also said that having ample funds is a separate matter from whether financing every investment with cash on hand is optimal. The company indicated it would comprehensively weigh investment needs across HBM, server DRAM, and enterprise SSDs, and compare the impact of internal and external funding on the economic value for existing shareholders.
It therefore cannot be said that SK hynix must raise profits at its Chinese plant in order to invest in the US. It is also a question of how to combine the use of existing facilities with new investment as funds are allocated across multiple growing businesses. As for Solidigm's use of capital, the statement said nothing had been decided at that point.
From the US government's perspective, allowing expansion at a China site in order to attract domestic production raises questions about consistency with the restrictive policy BIS has announced. On the other hand, tightly limiting near-term supply could also affect business decisions when companies invest in new plants. The negotiation report shows that the long-term goal of increasing manufacturing capacity in the US cannot be separated from the question of how to use existing plants until then.
The indicators of progress in the negotiations are the use and scope of equipment approved for Dalian, and concrete investment decisions on a US NAND plant. Only when both become clear can the plan be assessed as to whether it can launch NAND production in the US while maintaining near-term supply from existing sites.