Choi Tae-won's Quiet Bid to Build a Semiconductor Fab in Japan
SK Group is quietly exploring a semiconductor joint plant in Japan, a strategic move that could reshape the Korea-Japan-US chip alliance. The implications go far beyond real estate.
The Statement That Wasn’t a Statement
When Choi Tae-won, chairman of SK Group, stood before reporters in Sendai on August 31, he gave them almost nothing. A Japanese semiconductor joint factory, he said, was under active review. Details would come when the discussion concluded. That was it.
The restraint was itself a signal. Choi was fielding questions from Bloomberg and Kyodo journalists who had converged on the Korea-Japan Chamber of Commerce meeting specifically to chase this story. He didn’t confirm or deny a deal. He didn’t name a partner. He simply refused to rule anything out.
In the semiconductor industry, that kind of calibrated ambiguity is never accidental. It means SK Group is at a stage where premature disclosure could damage negotiations — or invite interference from Washington and Seoul.
Why Japan? Why Now?
The most obvious reason is cost. AI-driven demand for high-bandwidth memory, or HBM, has turned SK Hynix into one of the fastest-growing memory chip makers in the world. But SK’s domestic capacity in Korea is constrained by land scarcity, rising utility costs, and increasingly strict export controls on advanced memory technology.
Japan offers the opposite profile. Miyagi Prefecture, where the Korea-Japan chamber meeting took place, has been circling SK Hynix for months. According to the Nihon Keizai Shimbun, the prefecture is offering a 300,000-square-meter site in the Ohiramura industrial park near Sendai — land large enough for a major fab complex, with industrial water and power already in place, and within commuting distance of a university town that can supply engineers.
Miyagi Governor Yoshihiro Murai didn’t mince words during a press conference on August 26. There is “almost no place in the world” like it, he said.
That’s not just regional boosterism. Japan’s government has been aggressively court semiconductor investment since Tokyo committed roughly $9 billion in subsidies to attract Intel’s new fabs in Kumamoto. The race to fill those incentive programs is real, and Miyagi sees SK Hynix as a way to diversify beyond a single American anchor tenant.
The Kioxia Factor
There is a structural reason this particular gamble makes sense for SK that doesn’t apply to most other foreign chipmakers looking at Japan. SK Hynix already owns approximately 14 percent of Kioxia Holdings, Japan’s former Toshiba memory division and one of the world’s largest NAND flash producers. That stake gives SK a foothold in Japanese industrial policy conversations, access to Kioxia’s existing supply chain relationships, and a built-in domestic partner that understands both the technical and regulatory landscape.
A joint factory with Kioxia — or at least one that leverages the relationship — would be the most natural configuration. It would also be politically legible in Tokyo, where a Korea-Japan memory partnership carries fewer historical sensitivities than a pure Korean fabrication plant on Japanese soil.
The Washington Variable
Which brings us to the obstacle that isn’t mentioned in any Japanese press release.
The United States has been applying increasing pressure on South Korea to restrict the export of advanced semiconductor manufacturing equipment and design tools to China. Any new fab built by SK Hynix outside Korea — especially one producing HBM chips used in AI training — would fall under scrutiny from the Commerce Department and possibly the Office of the U.S. Trade Representative.
SK Hynix’s existing HBM production is concentrated in Korea, precisely because the technology is subject to tight U.S.-led export controls. Moving that production to Japan would require Washington’s implicit blessing, or at minimum its indifference. Neither is guaranteed.
Seoul adds another layer of complication. The South Korean government has its own concerns about technology leakage, particularly around the most advanced memory products. Any major overseas fab expansion requires approval from the Ministry of Trade, Industry and Energy, and that ministry will weigh the geopolitical implications carefully.
What This Means for TSMC’s Dominance
TSMC’s expansion in Kumamoto has been widely covered. What gets less attention is the parallel pressure building on SK Hynix to do the same — to locate advanced memory production closer to Japan’s materials and equipment supply chain, which remains the most concentrated in the world. Tokyo Big Three chemicals companies, Screen Holdings, ASML’s Japanese partners: the infrastructure for a memory fab exists in Japan in a way it doesn’t elsewhere in the region.
If SK Hynix builds a fab in Miyagi, it wouldn’t just be a cost play. It would be a strategic hedge against the growing concentration of advanced semiconductor manufacturing in Taiwan, and a signal that memory chipmakers — long seen as less strategically critical than logic fabs — are ready to compete for the same geopolitical tailwinds.
That’s the non-obvious implication: SK Group’s move, if it materializes, would blurring the line between the memory and logic halves of the chip industry in a way that benefits everyone except the companies that haven’t yet made the same calculation.
Who Wins, Who Loses
Miyagi Prefecture wins immediately — political capital, Jobs, a seat at the table in Japan’s semiconductor revival narrative. Kioxia wins if the partnership deepens, gaining access to SK’s HBM expertise and capital. SK Hynix wins if the fab comes online faster and cheaper than a Korean expansion.
Washington loses leverage over Seoul’s memory supply chain if the new fab operates under Japanese — rather than Korean — jurisdiction. Beijing loses an alternative routing for advanced memory technology if the new facility falls under U.S. export control frameworks. And TSMC loses a piece of the narrative that memory and logic are separate worlds, which is useful to no one.
What Happens Next
Choi Tae-won said the announcement would come “once discussions conclude.” That timeline is impossible to pin down. Semiconductor site negotiations in Japan typically run 18 to 36 months from initial接触 to groundbreaking, assuming no regulatory complications. Given the current political sensitivity, this one could take longer.
What’s clear is that SK Group is no longer just reacting to the AI chip boom from behind Korean borders. It is actively reshaping its geography, and Japan — with its land, its supply chain depth, and its appetite for foreign investment — is becoming a central part of that plan. The question is no longer whether SK will build in Japan. It’s whether it can do so without triggering a diplomatic incident it doesn’t need.