The bidding took place this week in a conference room rather than an auction house. Investment banks pitched for roles on what could be one of the largest American listings in years.
Reuters reported on September 25 that Solidigm, the U.S. NAND subsidiary of South Korea’s SK Hynix, is considering a share sale in the United States as early as 2027 at a valuation of as much as $150 billion. The company could raise roughly $15 billion, according to people familiar with the matter, though both the size and the timing remain subject to market conditions.
Solidigm is headquartered in Rancho Cordova, California, a suburb of Sacramento, and its history is one of resale. The business began inside Intel as its NAND and solid-state drive unit. SK Hynix agreed in 2020 to buy it for about $9 billion, and the operation began running under its own name in 2021.
The valuation being floated would put the subsidiary in a different category from the price its parent paid. Memory pricing explains most of that gap. NAND and DRAM contracts have climbed through a cycle driven by AI servers, which consume far more flash storage than the general-purpose machines they replaced.
Building data centers for model training requires enormous pools of fast storage to feed GPUs that would otherwise sit idle. Enterprise solid-state drives have become a bottleneck component, and suppliers have regained pricing power they had lost during years of oversupply.
SK Hynix has signaled that it wants a separate market price for that business. The parent said last month that it was reviewing options to improve Solidigm’s competitiveness, language that analysts read as preparation for some form of separation. Earlier reports said the subsidiary planned to raise about 5 trillion won, roughly $3.6 billion, ahead of any listing.
There is also a factory question attached to the story. Reuters reported in early September that Solidigm is weighing a new NAND plant in the United States, with upstate New York among the candidate sites. A domestic plant would fit the broader push by Washington to move advanced semiconductor manufacturing onshore, and it would give a U.S.-listed company a local narrative to sell.
The strategy follows the logic of the memory cycle itself. Memory makers are among the most volatile large businesses in technology, and their valuations swing with contract prices in a way that diversified chip companies’ shares do not. Listing the flash unit separately would let public investors pay for that exposure directly rather than through a conglomerate whose results are also shaped by DRAM, image sensors and packaging.
It would also create a second currency. SK Hynix has spent heavily to keep pace with Samsung Electronics and Micron Technology in high-bandwidth memory, the stacked DRAM that sits beside AI accelerators. A separately listed NAND arm gives the group another set of shares to use for acquisitions, employee compensation and joint ventures, at a moment when memory capital spending is climbing again.
Analysts said the mechanics of a $15 billion raise would make Solidigm one of the largest listings on a U.S. exchange in recent memory, which raises its own problem. Deals of that size need a deep and receptive market, and the window can close quickly. A date in 2027 gives the company room to wait but also leaves it exposed to whatever the memory cycle is doing when it finally files.
The competitive picture is unforgiving. Samsung, Micron, Kioxia and Solidigm itself all sell largely interchangeable flash, and the industry has a long history of adding capacity until prices collapse. AI demand has interrupted that pattern, not repealed it. Each of the four has announced capacity additions in the past year, and the standard industry error is to bring that capacity online just as demand normalizes.
That timing problem is the central risk in holding the asset rather than selling it. If Solidigm lists at the top of a pricing cycle and the cycle turns, the parent will have sold a minority stake in a business whose earnings peak arrived before its public history began. If it waits and prices keep climbing, the delay costs nothing but patience.
Of the several considerations in play, the simplest one may be competitive. A standalone Solidigm would compete for talent and customer attention against Samsung and Micron directly, with its own equity to pay people in. Companies inside conglomerates rarely lose that argument, and the case for carving out a flash business rests mostly on whether investors will eventually pay more for it alone.
For SK Hynix, the calculation is unresolved. Separating its most improved asset would crystallize value now, and it would also hand shareholders a cleaner scorecard to grade the parent against. Whether the company wants that comparison is a question it has not answered.
