SK's Nuclear Option With Kioxia Rewires Korea-Japan Chip Politics

The Threat Is the Product

Choi Tae-won, chairman of SK Group, told Japan’s Asahi Shimbun that his company is considering joint production with Kioxia—the largest Japanese maker of NAND flash memory. Then he added something most executives would leave unsaid: if the partnership doesn’t move forward, SK will walk away and terminate its entire investment in Kioxia.

The message was blunt. It came wrapped in diplomatic language, but the substance was a lever being yanked. SK holds convertible bonds in the special-purpose company that owns Kioxia after Toshiba divested its chip business last month. If Choi converts those bonds into shares, SK Hynix stands to gain controlling voting power over Kioxia’s board. The threat isn’t vague—it’s backed by a specific financial instrument sitting on a balance sheet, ready to flip an alliance or sever one entirely.

This changes the geometry of the conversation. Kioxia was negotiating from what it assumed was a position of scarcity—Japan was the only viable location for SK’s next fab. Choi has just made it clear that the scarcity runs both ways.

Why Kioxia Can’t Just Say No

Kioxia sits on a strategic asset that the rest of the industry needs: deep-trench NAND flash technology refined over decades at Toshiba’s Kumamoto facility. SK Hynix, the world’s second-largest memory chipmaker after Samsung, needs that capacity badly. Choi estimates a 20 to 30 percent shortfall in global data-center memory demand driven by the generative AI boom. SK’s domestic investments alone won’t close that gap.

Japan offers something Korea cannot replicate at scale. Its semiconductor ecosystem concentrates equipment manufacturers, materials suppliers, and fabrication know-how in tight geographic proximity. For an industry where a single contamination event can scrap an entire wafer lot, that proximity matters. Choi called it “affinity”—a cultural and industrial alignment between Korean and Japanese manufacturing disciplines that he argues makes Japan a uniquely suitable site for new capacity.

But Kioxia has its own headaches. The transition from Toshiba ownership to the American-backed SPC—essentially a vehicle controlled by investment funds—has left the company in a managerial limbo. Its largest shareholder structure is unclear. Governance questions are hanging over the board. Choi knows this. The timing of his public comments suggests he’s counting on Kioxia’s leadership being too fragile to absorb another major disruption.

The Convertible Bond as a Weapon

SK Hynix holds convertible bonds issued by the SPC that owns Kioxia. Upon conversion, those bonds would grant SK the majority of voting rights in the SPC—effectively making it the de facto controller of Kioxia’s strategic direction. That is the invisible hand behind Choi’s words.

If SK converts and takes control, it could force a production-sharing agreement, redirect Kioxia’s R&D priorities, or restructure the company’s output to prioritize SK Hynix’s AI-bound memory products. If it doesn’t convert, Kioxia remains independent but financially weakened, dependent on the very investor whose convertible bonds it issued.

Either outcome favors SK. That’s the design.

The Second-Order Shockwaves

What Choi has unleashed extends well beyond a bilateral negotiation. Memory chip markets are notoriously cyclical, and the current AI-driven demand surge has already pushed prices to levels not seen since 2022. A disruption to the SK-Kioxia supply arrangement would ripple through every downstream sector that depends on NAND flash: cloud providers, automotive manufacturers, consumer electronics firms, and defense contractors. Data-center operators bidding for GPU clusters now also need to secure memory inventory, and any hint that Kioxia’s output could be constrained by shareholder conflict would tighten an already agonizing supply equation.

There is also a geopolitical dimension that Tokyo cannot ignore. Japan’s semiconductor revival has been one of the Crown Jewel priorities of the Kishida and subsequent governments, backed by billions in subsidies tied to conditions that foreign investors respect Japanese operational autonomy. SK’s public pressure campaign effectively tests whether those conditions hold when a major Korean buyer decides it no longer needs to ask nicely. If SK converts its bonds and gains control, it would represent the first time a Japanese semiconductor company with strategic military applications—Kioxia supplies memory for Japan’s defense industrial base—falls under significant foreign voting dominance. That scenario would almost certainly trigger scrutiny from METI and potentially the Cabinet Secretariat’s screening mechanisms.

On the Korean side, Samsung is watching closely. The rivalry between SK Hynix and Samsung in memory chips is fierce, and any move that expands SK’s control over NAND capacity shifts the competitive balance. Samsung has been investing aggressively in its own NAND facilities in South Korea and planning expansions in Texas, but those projects carry long lead times and enormous capital requirements. SK’s ability to lock in Japanese capacity through financial leverage rather than greenfield construction gives it a timing advantage that could prove decisive through the remainder of this demand cycle.

Regional governments are not immune either. Several Japanese municipalities—including those in Kyushu and the Kanto plain—have been courting SK with land deals, tax breaks, and streamlined permitting. Choi confirmed he has heard from multiple cities. That competition among local governments creates a subtle internal pressure within Japan: the central government wants strategic coordination, but municipalities eager for jobs and investment may offer terms that undercut Tokyo’s broader industrial policy aims. It also gives SK additional leverage—if Kioxia or METI pushes back too hard, SK can signal that it has alternative sites ready.

Who Wins, Who Loses

SK Hynix wins first and foremost. It gains leverage over a critical supply node without committing upfront capital to build a new facility from scratch. It keeps its options open—if Kioxia cooperates, SK shares the fab; if not, SK walks and the bondholder claim still pressures Kioxia into negotiations.

Japan loses something subtle but real. The assumption that Japanese semiconductor assets would attract foreign investment on Japan’s terms is now under challenge. SK is effectively saying: we will build in Japan, but we will set the terms. That reverses the traditional dynamic where Japan’s ecosystem acted as the gatekeeper.

Kioxia’s employees and suppliers face the most immediate uncertainty. A change in controlling shareholder or a forced production-sharing arrangement could reshape shifts, supplier contracts, and hiring plans at the Kumamoto and Yokkaichi plants. No timeline has been given, but Choi signaled SK could announce its investment decision before year’s end.

What Comes Next

Choi has heard from multiple Japanese municipalities seeking SK’s investment. That creates competitive pressure on the central government too. If a local government offers better incentives, SK may have a fallback option that weakens Kioxia’s negotiating position further.

Kioxia’s response has been characteristically noncommittal—its PR office pointed to existing MRAM research collaboration with SK Hynix and noted that SK is one of its DRAM suppliers, expressing a desire to “continue maintaining good relations.” Translation: we see the threat, we won’t concede publicly, but we’ll keep talking.

The longer-term implication cuts wider than any single deal. The Korea-Japan semiconductor axis is being rewritten in real time, driven not by government policy but by corporate maneuvering around a memory shortage that the AI boom has made acute. The players who adapt fastest to this new leverage dynamic—not the ones with the most factories—will shape where the next generation of chips comes from.

SK has made its opening move. The question is whether Kioxia has the leverage left to respond—or whether it will be forced to accept terms that redefine what Japanese semiconductor sovereignty means in practice.