Samsung and SK Hynix Going Public in the U.S.: The Storage Bubble Is Sounding the Alarm

SK Hynix has officially filed an ADR listing application with the U.S. Securities and Exchange Commission this week, followed closely by Samsung Electronics management hinting at a “capital structure optimization” plan at its shareholder meeting.

Meanwhile, spot prices for memory chips have just hit an all-time high, HBM4 capacity pre-sales have sold out, and the numbers in the financial report look dizzyingly impressive.

However, even as the industry turns in what it calls the “best results ever,” it is also rushing to raise funds and expand capacity—when the cycle has often already reached its most dangerous point.

This is not only a routine move in capital operations, but also an industry-cycle rule responding with an emergency stress reaction at a critical moment.

In the narrative where AI compute demand is amplified without limit, it seems we have forgotten the semiconductor industry’s bloodiest lesson: prosperity often breeds excess, and financing is often the last alarm bell before the celebration ends.

• The paradox at the profit peak: why are cash cows suddenly rushing to “find money”?

In today’s memory industry, it is in a seemingly unassailable spotlight phase. The 2026 Q1 earnings reports show that SK Hynix and Samsung Electronics’ operating profits have both surpassed historical highs, and their gross margins even exceeded those of some logical chip design companies.

Greedy demand from AI servers for HBM (high-bandwidth memory), along with AI phones and AI PCs driving DDR5, has transformed memory chips from past cyclical commodity-like products into “strategic scarce resources.” With prices and profits rising in tandem, memory chips have become the most certain “hard asset” in this round of the tech rally.

But at precisely this point in time, the industry has taken an action that is extremely counterintuitive—on one side, profits have reached historical highs and operating cash flow is sufficient to cover day-to-day expenses; on the other side, it is rushing to list in the U.S. via ADR, bring in more capital, and push for capacity expansion.

SK Hynix’s fundraising in the U.S. is expected to reach as much as 10 billion dollars, and although Samsung Electronics has not made an explicit statement, the market widely expects it to follow with the issuance of global depository receipts.

On the surface, this is a natural choice for “valuation repair.” Korea’s stock market has long had a “Korea discount,” and limited liquidity constrains valuations. Compared with U.S.-listed peers like Micron Technology, Samsung Electronics and SK Hynix have long had P/E ratios suppressed at low levels.

By using ADRs to tap a pool of U.S. dollars, the move is essentially a “reset of pricing,” allowing global investors to buy these core Asian assets directly with dollars. It sounds reasonable and aligns with the instinct of capital chasing returns.

But if you place this behavior within a cycle framework, it takes on a completely different meaning. In the long history of the semiconductor industry, when companies do not lack money but choose to raise funds proactively, it often means they believe the current window is the best time to expand capacity.

Management is more aware than investors of the cycle of capacity construction: the money raised now is to buy equipment delivered next year, and it is to release capacity the year after that. And this typically happens near the top of the cycle.

Historically, whether it was the memory boom cycle of 2017–2018 or the earlier early stage of the 2010 smartphone explosion, industry leaders almost always massively lever up and expand capacity during the phase when profits are at their best. In 2017, when DRAM prices surged, the three major players all announced capital expenditure plans on the order of hundreds of billions of dollars, and the media cheered that a “semiconductor super cycle” was coming.

Yet that is precisely what becomes the starting point for a supply-demand reversal. When capacity finally concentrates release in the second half of 2018, demand slowed due to trade frictions and smartphone market saturation, and prices subsequently collapsed.

Today’s scenario is all too similar: the AI narrative provides the perfect reason to expand capacity, and abundant profits provide the strongest backing for expansion. When the leaders stop being cautious and start rushing to grab future market share, the dangerous seeds are already planted.

• The capital’s coded message behind ADR: a signal amplifier for supply expansion

The market is currently more willing to interpret ADRs as a positive catalyst, treating them as a channel for foreign capital into the Korean stock market. From an investment perspective, the real significance of this move is that it reinforces the certainty of future supply expansion.

Take TSMC as an example: its ADR listing has indeed brought long-term valuation improvement, but it is also accompanied by ongoing large-scale capital expenditures and capacity expansion. This model works logically because advanced process technologies have extremely high barriers to entry, strong customer stickiness, and capacity expansion can often be locked in by long-term orders. But the memory industry is fundamentally different.

Memory chips are a typical cyclical product. Their prices depend heavily on the relationship between supply and demand. Once the industry collectively expands capacity, price elasticity flips quickly. In other words, capital expenditure itself is a “reflexive variable” of the cycle: the more upbeat it is, the more expansion happens; the more expansion happens, the more likely it is to trigger the downturn cycle.

Memory chips do not have very high customer barriers. Samsung’s products and Micron’s products are highly homogenized at the commodity level, and competition hinges on cost and scale.

When SK Hynix fires the first shot with its ADR, and Samsung Electronics is also pushed to follow, the market needs to reinterpret this signal: this is not just “valuation arbitrage” in capital markets, but also “forward-looking supply” at the industry level.

The direct consequence of successful financing is that these giants obtain cheaper U.S. dollar capital to buy ASML lithography machines and to buy Applied Materials’ deposition equipment. In the equipment delivery cycle of 2026, orders signed today imply newly added wafer capacity 18 months later.

More importantly, driven by the AI narrative, the market has priced storage demand expectations extremely optimistically. Investors believe AI models will grow infinitely large and data will increase infinitely, so storage demand is rigid. But this expectation overlooks efficiency gains brought by model optimization.

As inference cost pressures rise, the software layer is aggressively reducing reliance on hardware. Once demand slows at the margin while supply accelerates its release, the price system will face dramatic adjustments. The huge pool of funds brought by ADRs will become the final straw that breaks the supply-demand balance. It enables expansion plans that might otherwise have been delayed due to cash flow constraints to be carried out smoothly, thereby locking in future oversupply in advance.

• The classic script at the cycle peak: from “the strongest ever” to “price collapse”

If we look back through history, memory industry cycle fluctuations have an extremely strong tendency to repeat, as if it were an inescapable fate.

At the previous peak in 2018, DRAM prices fell from their highs by more than 50% within just a single year, and the stock prices of industry leaders generally retreated by 50%–70%; in the more extreme 2008 cycle, the drop could even reach 70%–80%. The root cause of such violent volatility is this: memory is not a “demand-driven industry,” but a “supply-driven industry.”

When the market is at a cyclical high, companies often make expansion decisions based on current price levels, but capacity release has a time lag. Once capacity is concentrated and lands, it rapidly breaks the supply-demand balance. The danger signal this time is that the AI narrative overlays the cycle, making market expectations for demand even more consistent and optimistic—which in turn amplifies the downside room.

The 2026 market environment is significantly different from 2018 in one key respect: the hidden nature of inventory. Under the cover of AI servers, channel inventories are quietly accumulating. Many vendors, in order to secure HBM capacity, have been forced to bundle purchases of large volumes of standard DRAM and NAND Flash. These inventories are not fully absorbed by end-market consumption; instead, they are piled up in the middle segments of the supply chain. Once end-market sales fall short of expectations, inventory drawdown pressure will instantly transmit upstream.

From the perspective of investment trend indicators, a typical cycle top often has three major characteristics: the best performance, the strongest expansion impulse, and the most optimistic market expectations. Today’s memory industry has all three points almost at the same time.

In earnings call transcripts, analysts are still asking whether “capacity is sufficient,” rather than whether “demand is sustainable.” In brokerage research reports, target prices are repeatedly raised, yet the diminishing risk of capital expenditure returns (ROIC) is rarely mentioned.

Even more worth worrying about is the involvement of financial leverage. Although the low-interest-rate environment in 2026 has ended, prosperity in the derivatives market has attracted large amounts of speculative capital into the memory sector. When spot prices show slight slackening, the “kill everyone” dynamic of highly leveraged capital will accelerate the price collapse.

History may not simply repeat itself, but it always sticks to the same rhyme. When everyone believes this time is different, it is usually the most dangerous moment.

• The trap of prosperity: when a capital binge meets physical limits

What ADRs may bring is perhaps short-term valuation repair, but in the long run, it is more likely to become an amplifier of supply imbalance. When the “strongest fundamentals” and the “biggest expansion impulse” appear at the same time, what the market truly needs to think about is not how much upside is left, but how quickly and how deeply the market will clear as this cycle falls back.

Behind this is an eternal proposition about human nature and the cycle. In boom periods, people tend to extrapolate linearly, believing today’s high prices are tomorrow’s bottom prices; in downturn periods, people tend to become overly pessimistic, believing demand will disappear forever. The harshness of the memory industry is that its physical attributes mean capacity cannot be switched on and off instantly, and its capital attributes mean expansion cannot be stopped instantly either. When a capital binge hits physical limits, oversupply is the only endpoint.

A massive amount of news and precise analysis—right here on the Sina Finance app