NextFin News - SK Hynix is weighing a US initial public offering for its NAND memory unit Solidigm as soon as 2027, at a valuation that could reach $150 billion, according to people familiar with the matter — a move that would turn the world's most valuable private chip asset into a public bet on whether the AI-driven memory squeeze is a structural re-rating or the top of a cycle.
The report, dated September 25, 2026, says the world's second-biggest memory maker is in talks with potential advisers about a listing for the NAND flash unit. A separate wire account put the potential valuation at as much as $150 billion for a 2027 flotation. SK Hynix has not confirmed any plan; the company has said Solidigm is considering several ways to strengthen its competitiveness and that nothing is decided, promising further disclosure when details become concrete or within three months.
The timing matters because SK Hynix only just completed its own US listing. On July 10, 2026, the parent company priced 177.9 million American depositary shares at $149 each, raising approximately $26.5 billion on the Nasdaq under the ticker SKHY — the largest-ever US share sale by a non-American company, ahead of Alibaba's $21.8 billion New York debut in 2014. The ADRs closed their first day at $168.01, a roughly 13% pop. Less than three months later, the group is reportedly sizing up a second, far larger US capital-markets event.
The stakes are immediate. A $150 billion valuation would make Solidigm one of the most valuable chip companies in the world and the biggest US listing since SpaceX's record $75 billion offering in June 2026. It would also hand SK Hynix a multi-fold return on its $9 billion purchase of Intel's NAND and SSD business, agreed in October 2020 and fully closed only in March 2025, when the final $1.9 billion payment transferred Intel's remaining NAND intellectual property and engineering teams to Solidigm. But it would ask public investors to buy a pure-play NAND asset at the moment memory prices are rising at a pace the industry has rarely seen — and at the moment that pace is most likely to slow.
Why list now: the valuation window is open, but it is a window
The logic of a 2027 flotation is straightforward: sell the asset while the market is paying for AI scarcity. NAND contract prices jumped an estimated 55% to 60% quarter-over-quarter in the first quarter of 2026, and enterprise solid-state drives have become the fastest-growing corner of the memory market. A research survey of the five largest enterprise SSD brands put their combined revenue at nearly $37.59 billion in the second quarter of 2026, more than double the prior quarter. Another research firm estimated enterprise SSDs at 48% of NAND bits shipped in the same period, up from 26% a year earlier. Solidigm, which controls roughly a quarter of the data-center-grade SSD market, is sitting on exactly the product mix investors are chasing.
SK Hynix's own numbers show how fast the tide turned. In the second quarter of 2026, the company reported NAND average selling prices up in the mid-50% range quarter-over-quarter, and Solidigm's high-capacity SSD revenue tripled. The group's NAND revenue, combining SK Hynix and Solidigm, reached about $7.53 billion in the first quarter of 2026, up 44.6% sequentially, with a 17.6% global NAND share — second only to Samsung's 31.6%. In that environment, a unit that lost money for years can be marketed as the cleanest listed proxy for AI storage demand.
There is also a balance-sheet logic on the parent side. SK Hynix crossed a $1 trillion market capitalization in May 2026 and briefly overtook Samsung Electronics as South Korea's most valuable listed company in June, at 2,080.4 trillion won ($1.35 trillion). The group has committed as much as $10 billion of investment into Solidigm's parent and is building a high-bandwidth memory fabrication plant in Yongin, South Korea. Monetizing a minority stake in Solidigm would recycle capital toward those priorities while crystallizing gains on an asset that, on paper, has already appreciated many times over its purchase price.
The balance-sheet problem beneath the supercycle
Any investor buying Solidigm in a public offering will inherit a balance sheet built during the worst part of the memory downturn. The unit posted cumulative net losses of nearly 8 trillion won from 2021 to 2023, turned its first annual profit of about 830.7 billion won in 2024, and carried a debt ratio of 4,484.6% last year — roughly 22 times the roughly 200% level generally considered healthy. Shareholder equity was negative 906 billion won in the first half of 2024, a state of capital impairment. SK Hynix had lent Solidigm about 11.32 trillion won of operating funds by early 2025.
That is the real arithmetic behind the reported pre-IPO fundraising of 5 trillion to 10 trillion won ($3.6 billion to $7.2 billion), with Morgan Stanley and Goldman Sachs named as lead managers and sovereign wealth funds, including Mubadala, sounding out a stake. A listing at the previously discussed 50 trillion won valuation — about $35 billion to $36 billion — would be one thing; a listing near $150 billion would be another. The gap between those two numbers is not just negotiation room. It is the difference between pricing the asset on its post-acquisition cleanup and pricing it as if the current NAND supercycle is a permanent regime.
There is a second, operational overhang: Solidigm's NAND fab in Dalian, China. The plant, inherited from Intel, is aging, and SK Hynix has been converting lines there to 192-layer NAND while planning a new 238-layer line at a separate facility with a phased ramp through the first half of 2027. In a public filing, that geographic and technological transition would be disclosed, priced, and debated every quarter.
Cyclical windfall or structural re-rating: the call
This is the question the $150 billion figure answers by assumption. The evidence says the current leg is cyclical, and the re-rating is only partly structural.
The cyclical case is overwhelming. Memory has never escaped its boom-and-bust cycle. SK Hynix swung from a record 23.5 trillion won operating profit in 2024 to the current windfall on the back of the same mechanism that has driven the industry for four decades: a supply shortfall meets inelastic demand, prices spike, capex follows, and capacity arrives. Intel's chief executive warned in 2026 that memory prices had surged more than 500% and could remain tight into 2027 — a statement that describes the peak of a cycle as clearly as any industry metric. Jing Jie Yu, an equity analyst at Morningstar, put it more directly: memory pricing power "only exist because of such deep undersupply," and as new supply comes online in 2027 and 2028, "supply/demand dynamics should improve tremendously, eroding the pricing power of memory players and leading to a cycle turn."
Memory pricing power only exist[s] because of such deep undersupply. As that supply comes online in 2027 and 2028, supply/demand dynamics should improve tremendously, eroding the pricing power of memory players and leading to a cycle turn.
The structural case is real but narrower. AI workloads do change the demand curve for storage: a single high-end AI GPU can require around 16TB of TLC or QLC NAND, and an AI server rack can demand more than 1,000TB. Enterprise SSDs moving from 26% to 48% of NAND bits shipped in a year is a mix shift, not a weather pattern. That mix favors Solidigm specifically, because it exited the consumer SSD market to concentrate on data-center products, and because its Intel heritage gives it depth in enterprise controllers and firmware. If AI storage intensity keeps compounding faster than capacity additions, the trough of the next cycle sits higher than the last one.
But a higher trough is not the same as an end to the cycle. The structural premium belongs to the companies with pricing power that survives oversupply — and NAND, the most standardized and capacity-driven of the memory markets, has the least of it. The right read: Solidigm is a structurally better-positioned company riding a cyclical price spike. Pricing it at $150 billion assumes the spike is the new floor.
The counter-thesis, and what would prove it right
The strongest argument against that skepticism is that this time the supply response is constrained. Production of high-bandwidth memory for AI accelerators consumes several times the wafer area of conventional memory per bit, and manufacturers have prioritized the most profitable AI products over volume growth. If capacity cannot ramp as fast as it did in the 2017-2018 cycle, the upswing lasts longer and the peak is higher. In that scenario, a 2027 listing lands not at the top but on a plateau, and $150 billion looks cheap in hindsight. SK Hynix's own HBM capacity is effectively sold out, and long-term agreements with hyperscalers have locked up portions of future output — exactly the visibility a public-market investor pays a premium for.
That argument is coherent, but it rests on one testable assumption: that supply growth stays behind demand growth through the 2027-2028 window. The falsifying signal is concrete. If NAND contract prices fail to rise quarter-over-quarter for two consecutive quarters after mid-2027 — or, more sharply, if enterprise SSD contract pricing prints flat to down while SK Hynix's and Solidigm's announced capacity expansions are still ramping — the cycle-turn thesis is confirmed and the $150 billion valuation premise breaks. A second signal: if SK Hynix's NAND average selling price growth decelerates from the mid-50% quarter-over-quarter pace reported in the second quarter of 2026 to single digits before Solidigm prices its offering, the window has already closed.
Who wins, who is exposed, and what to watch
The immediate beneficiaries of a Solidigm IPO are clear. SK Hynix would crystallize a multibillion-dollar gain on a $9 billion acquisition that looked questionable for most of its first three years, and it would fund its HBM and US fab buildout without further diluting its own shareholders. US investors would finally get a pure-play, dollar-denominated bet on AI storage, a niche that has been accessible only through diversified Korean and Japanese parents. Lead underwriters and the pre-IPO consortium would capture fees on what could be the largest tech flotation since SpaceX.
The exposed are the buyers at the offer price. A $150 billion entry multiple for a NAND asset pricing in sustained mid-50% quarterly price gains is a bet that the cycle has been repealed. History suggests otherwise: the same analysts who call this a supercycle also date its turnover to 2027-2028. The asymmetry is simple — SK Hynix sells into strength, and public shareholders inherit the downside when capacity catches up.
Three scenarios frame the next 12 months. In the base case, SK Hynix keeps the plan in the "considering" stage through 2026, completes the 5 trillion to 10 trillion won pre-IPO round at a valuation closer to the 50 trillion won figure already discussed, and waits for the 2027 earnings visibility before pricing a public offering — likely well below $150 billion. In the upside case for sellers, NAND prices keep climbing through 2027, hyperscaler capital spending shows no sign of slowing, and a late-2027 listing clears $100 billion as the biggest chip IPO in history. In the downside case, memory prices flatten in the first half of 2027, the debt ratio and Dalian overhang dominate the roadshow, and the IPO is delayed or restructured into a smaller stake sale.
The short-term read is bullish for SK Hynix's own shares: the mere prospect of a Solidigm valuation pop adds a free option to a stock that has more than doubled this year. The medium-term read is neutral: the parent's fortunes still turn on DRAM and HBM pricing, not on a storage subsidiary's multiple. The long-term read is the one that matters for IPO investors: Solidigm is a better company than the one SK Hynix bought from Intel, but it is still a NAND company, and NAND has never rewarded investors for believing a shortage is permanent.
The market will get the Solidigm listing it deserves — the question is whether it gets it at $35 billion or $150 billion, and that gap is the entire difference between a fair deal and a cycle-top transfer.
Explore more exclusive insights at nextfin.ai.
