#SK Hynix Resumes Dalian Factory# [A timeline that’s easy to overlook, revealing the game currently unfolding in the global semiconductor industry] "The U.S. pressures it to build a factory, but SK Hynix resumes its Dalian operations first."
The equipment permit the U.S. issued to SK Hynix’s China plant covers only through the end of this year. According to reports from Korean media, the Dalian NAND plant’s operations, which have been suspended for four years, will—at the earliest—move production equipment in by November. Ten days after this news appeared, SK Hynix also reportedly said it is preparing to build a large memory chip factory in Miyagi Prefecture, Japan, with a potential investment amounting to tens of trillions of South Korean won.
On one hand, SK hynix is rushing to restart the Dalian production lines before the annual licenses expire, while on the other hand it is assessing whether to locate the next factory in Japan. SK hynix appears to be expanding capacity globally, but behind the scenes it is a direct collision between two forces: China has factories, workers, and supply chains that can be put into production immediately—production bases that SK hynix cannot bypass if it wants to continue expanding its global market. Meanwhile, the U.S. is trying to use equipment licensing to control SK hynix and keep future advanced production capacity within the allied system, ideally within the U.S. itself.
Media reports from South Korea indicate that SK hynix plans to invest tens of trillions of won to build a memory-semiconductor manufacturing plant in Miyagi Prefecture in Japan’s northeastern region. Note that over the past few years, South Korea and Japan have had continuous frictions over chips and semiconductors, and have been keeping each other in check. This would be the first time a South Korean semiconductor company is investing in semiconductors in Japan and establishing a production base.
SK hynix’s decision to build a factory in Japan is partly because Japan provides large subsidies for foreign companies to build factories. Reports show that Japan has already provided up to 476 billion yen in support for TSMC’s Kumamoto project, and 500 billion yen to support Micron’s factory construction in the U.S. If SK hynix’s project in Japan moves forward, the Japanese government would also take on part of the factory construction costs.
Another factor is Japan’s advantages in materials and equipment. Japan still leads in areas such as silicon wafers, photoresists, packaging materials, and various semiconductor equipment. SK hynix’s chip production lines already use materials and equipment from Japanese companies such as Namics, Shin-Etsu Chemical, Tokyo Ohka, and others. Building factories in Japan can directly embed upstream suppliers into the supply chain.
However, if SK hynix is going to start production in Japan, on the one hand it will definitely face domestic public opinion in South Korea; on the other hand, the United States will also have complaints.
The U.S. has been putting pressure on the South Korean government and companies, demanding that they invest in and build memory-chip factories in the U.S. According to a report from a North Korean media outlet, the U.S. has already asked South Korean semiconductor companies to build memory-chip factories in the U.S., and it has also expressed dissatisfaction with South Korea’s announcement of an 800 trillion won Korean semiconductor cluster in South Chungcheong/Hunan? (as reported: the Korean province of South Jeolla). The U.S. believes South Korea has prioritized keeping the next round of the most scarce memory production capacity and huge capital within the country, yet it has been slow to come up with a plan to build memory factories in the U.S.
While the U.S. and Japan are secretly building up strength to compete for South Korea’s storage-factory business, SK hynix’s Dalian plant in China is already close to completing its capacity expansion.
SK hynix’s subsidiary, Solidigm, has resumed investment in the Dalian Plant 2 in the first half of this year. The company aims to have equipment moved in as early as November this year, and to achieve mass production in the first half of 2027. The added monthly die-attachment capacity is about 40,000–60,000 wafers. Dalian Plant 1 currently has a monthly production capacity of about 100,000 wafers. After Plant 2 goes into operation, the total monthly capacity of the Dalian base will reach 150,000 wafers, approaching nearly half of its NAND capacity. Moreover, if you look only at overseas NAND front-end manufacturing, Dalian is currently SK hynix’s only production base, accounting for 100%.
It should be noted that SK hynix’s Dalian Plant 2 is not an entirely new project. SK hynix took over Intel’s Dalian NAND factory in 2021. In May 2022, it started construction of Plant 2, but then operations were halted due to a sharp collapse in NAND prices and U.S. equipment export controls. With a surge in demand for memory chips over the past two years, SK hynix restarted the project in the first half of this year.
As for why SK hynix is restarting the Dalian Phase II project right now: on one hand, everyone can see the memory-chip price hikes—SK hynix wants to lock in capacity by expanding production bases. On the other hand, there is also a more隐蔽 reason.
Last August, the U.S. Department of Commerce revoked the existing “end-user verification” qualification for Samsung and SK hynix’s China plants. By the end of the year, the U.S. again issued equipment import licenses for the 2026 fiscal year to Samsung and SK hynix, allowing them to continue shipping some U.S. equipment to their China factories. In other words, by this coming year-end, SK hynix’s Dalian Plant 2 is unlikely to face too much pressure regarding advanced equipment imports.
Placing the Dalian project within SK hynix’s current global investment map makes the significance of SK hynix’s probing even more obvious.
In South Korea, SK hynix has already approved an investment of 54.3 trillion won to build two plants: the REni Y2 plant and the Cheongju M17 plant. The first cleanroom at Cheongju M17 is expected to be completed by the end of 2028, while the first cleanroom at Icheon Y2 will not be completed until mid-2029. In the United States, SK hynix has already determined an investment of $3.87 billion to build an HBM advanced packaging and R&D base in Indiana, with operations planned for at least 2028.
That is to say, even if capacity expansion proceeds as fast as possible, during this long “vacuum period” from this year to next year, SK hynix can still only choose China’s supply chain to meet production capacity. This is determined by China’s industrial structure. So if the U.S. wants to control SK hynix and make it circumvent China, it also has to consider what it will use to satisfy the customers’ ever-growing orders.
This also exposes the inherent contradictions in the U.S. chip export controls. The U.S. hopes to use equipment licensing to restrict China’s factories from expanding production capacity and upgrading technology, and to pull SK hynix’s next round of investment toward the United States. But at the same time, AI companies in the U.S., South Korea, and Japan need more HBM, DRAM, and NAND, and cannot afford for China’s existing memory capacity to suddenly exit the market. Moreover, South Korean companies will not completely follow U.S. instructions—leaving a great business opportunity on the table would make no sense.
SK hynix’s three-pronged approach appears to be fortunate on all sides, but in reality it is a company’s difficult survival between a political iron curtain and the laws of industrial development. It cannot abandon the readily available plants and supply chains in China, cannot refuse Japan’s material advantages and huge subsidies, and also does not dare to ignore the market and political pressure from the United States. This entanglement of “nothing can be left out” precisely confirms that over decades of semiconductor globalization, the supply-chain network woven across the world has never been something that can be dismantled by administrative orders alone.
Geopolitics may delay the construction of factories, but it cannot rewrite the physical laws of industry. In this game over memory chips, the final winner will not be the party with the loudest pressure tactics, but the one that can truly get lithography tools to run.
