Kioxia, the Japanese memory-chip maker, is considering listing in the United States rather than on the Tokyo exchange where its shares already trade. Multiple Chinese financial outlets reported on September 14, citing people familiar with the matter, that the company is weighing a U.S. offering that could raise about $10 billion. Kioxia has held discussions with Bank of America, Goldman Sachs, and JPMorgan, the reports said.
The news landed badly in Tokyo. Kioxia’s shares fell 6.5 percent on the day, a sign that investors read a U.S. listing as dilution or as a signal that the company wants access to deeper pools of capital than its home market can offer. The company has not commented publicly on the reports.
Kioxia was carved out of Toshiba, one of Japan’s oldest industrial names, and is one of the world’s leading suppliers of NAND flash memory. Its chips store data in everything from smartphones to data centers, and the company sits near the top of a market dominated by a handful of Asian producers, including Samsung and SK Hynix.
A U.S. listing would put Kioxia alongside the chipmakers that have ridden the artificial-intelligence boom to rich valuations in American markets. NAND demand has been lifted by AI servers, which require large amounts of storage, and by the same data-center buildouts that have pushed memory prices higher.
The company has tried to list before. An IPO planned for 2020 was shelved after the pandemic and U.S.-China trade tensions clouded the outlook, and the company has restructured repeatedly as its private-equity backers sought an exit. A fresh attempt, this time in New York, would signal that the market window has reopened.
The proposed $10 billion raise would rank among the largest listings of a Japanese technology company in recent years. It would also test whether U.S. investors, who have grown selective about new offerings, will pay up for a memory maker whose fortunes swing with the industry’s price cycle.
NAND prices are cyclical in a way that AI accelerator demand is not. When supply exceeds demand, prices fall sharply and profits compress; when demand outstrips supply, the same companies report record results. Analysts said investors will weigh that volatility against the current upswing.
The move comes as memory makers are spending heavily on capacity. Samsung and SK Hynix have announced large investments in new fabrication plants, and Kioxia faces the same pressure to keep up with technology transitions. A U.S. listing would give it capital to fund that expansion.
Bain Capital led the consortium that bought Toshiba’s memory unit in 2018 for roughly $18 billion, after Toshiba’s nuclear-construction losses forced it to sell its most valuable asset. The deal left Kioxia in private hands for years, through a period when the memory market boomed and then collapsed during the pandemic and its aftermath.
The company’s ownership structure adds a layer of complexity. The Bain-led consortium has been looking for an exit for years, and a U.S. IPO would begin to unlock that stake. Skeptics noted that Kioxia already has a listing in Japan, and a secondary U.S. offering is more unusual than a pure IPO. The company could choose to issue American depositary receipts rather than a full primary listing, and the structure will matter to how investors value the shares, people familiar with the discussions said.
NAND and DRAM are different businesses, but the AI buildout has lifted both. DRAM handles the fast, short-term memory that accelerators need, while NAND provides the persistent storage that holds the enormous datasets used to train models. Data centers buy both, and the industry’s leaders have reported that AI demand is absorbing a growing share of output.
The proposed listing would also place Kioxia in direct comparison with Micron Technology, the Idaho-based maker that is the only major U.S. memory producer. Micron’s shares have risen sharply on AI demand, and analysts said Kioxia’s bankers would likely frame the Japanese company against that benchmark when marketing the deal.
The timing is not accidental. Memory prices have been rising through 2026, and the same week Kioxia’s listing talks surfaced, the broader market was wrestling with whether the AI buildout is moving too fast. A memory maker seeking a rich valuation is, in effect, betting that the cycle still has room to run.
NAND flash is the storage workhorse of the AI era, and its makers have moved from boom-and-bust toward a steadier demand profile as data centers order years ahead. That shift is part of what a U.S. listing would sell to investors: a memory business less exposed to the old consumer cycle and more tied to the durable buildout of AI infrastructure.
For now, the reports describe early-stage conversations. Banks have been contacted, but no deal has been announced, and terms could change. The company’s silence leaves investors to read the Tokyo selloff as the market’s first verdict on a listing that has not yet happened.
