Track Hyper | SK Hynix bets on QLC flash memory to reshape the storage cycle

Author: Zhou Yuan / Wall Street News

In the cyclical nature of the semiconductor industry, memory chips have always been the most volatile segment.

SK Hynix announced in late August that it has achieved mass production of 321-layer 2Tb QLC (Quad-Level Cell) NAND flash memory and plans to enter the AI data center market.

This is not only a technological breakthrough for SK Hynix but also a strategic move in the capital logic of industrial competition.

From an investor's perspective, this product iteration may signal the beginning of a new storage cycle.

The Cyclical Fate of the Storage Industry

The memory chip industry has always exhibited a distinct "Bullwhip Effect."

In the book "The Fifth Discipline," a classic case of demand fluctuation is described: minor changes on the demand side, due to information asymmetry in the supply chain, are amplified upstream into severe inventory and price fluctuations.

Taking NAND flash memory as an example, over the past decade, it has repeatedly experienced cycles of "oversupply leading to price crashes, manufacturers suffering losses and cutting production, followed by price surges due to demand recovery."

This cyclicality makes investors both wary and hopeful about the storage industry.

On one hand, product prices are highly sensitive to profitability; on the other hand, each new trend on the demand side (smartphones, cloud computing, AI) could become a catalyst for reversing the cycle.

SK Hynix's decision to launch 321-layer QLC NAND at this time clearly bets on the structural demand brought by the AI wave.

Currently, AI has become the largest incremental demand source in the global semiconductor industry.

Whether for training large models or inference services, massive data storage is required behind the scenes. Unlike the consumer market, the procurement logic of AI data centers places greater emphasis on the balance of capacity, performance, and cost.

QLC NAND precisely meets this demand: although its write endurance is not as good as TLC (Triple-Level Cell), its high capacity and low-cost advantages are more pronounced in read-dominant scenarios.

SK Hynix positions the 321-layer QLC for AI data centers rather than traditional consumer applications, indicating that its business logic has shifted from supply-driven to demand-driven.

For the capital market, this is a signal: AI may become a new pivot for the NAND cycle in the coming years. In fact, this is no longer a potential signal but a market consensus that has already been reached.

In the global NAND market, Samsung, SK Hynix, Micron, and Kioxia have long held a monopoly.

After entering the 300-layer demand phase, subtle changes in the landscape have emerged.

Samsung still holds pricing power due to its production capacity, Micron emphasizes differentiated processes, Kioxia and Western Digital are striving to catch up; while SK Hynix chooses to break through with QLC, directly entering the AI data center market.

From an investment logic perspective, whoever can first secure large-scale AI clients may gain pricing power in the next 2-3 years.

This is because the growth certainty of the AI market is far higher than that of traditional consumer electronics, and once long-term supply agreements are signed, it can mitigate the risks brought by price fluctuations.

From this perspective, the 321-layer QLC is not just a product but also a bargaining chip for SK Hynix to compete for pricing power The profitability of storage manufacturers largely depends on whether the cost curve can decline faster than the price drop curve.

321 layers mean higher storage density, which can significantly dilute the unit bit cost if the yield is stable. For investors, this directly determines whether the company can maintain positive profits during a price downturn.

It is worth noting that SK Hynix's profitability in the NAND business over the past two years has not been impressive, and at one point, it even suffered losses due to price declines.

This bet on QLC is not only a leading technological attempt by SK Hynix but also a strategic choice to seek a turning point for profit recovery.

If the demand in the AI market materializes, the gross margin of SK Hynix's NAND department is expected to improve significantly.

Capital Expenditure and Capacity Layout Game

Another core variable in the semiconductor industry is capital expenditure.

Over-investment often leads to oversupply, dragging down prices, while under-investment may result in missed opportunities during demand surges.

This is a dilemma. Achieving balance tests the deep insight of various storage companies into market trends and demand.

SK Hynix's decision to launch high-layer QLC NAND amid a general contraction in global capital expenditure is, in fact, a counter-cyclical layout.

The strategic intent here is to seize technological advantages while gaining a relative first-mover advantage during the industry's cautious overall capacity expansion.

For the capital market, this "expansion in the trough" strategy is often a focal point for counter-cyclical investors.

Data center customers differ from the consumer market; they tend to establish long-term supply relationships with upstream suppliers to ensure stable supply and controllable costs.

If the 321-layer QLC is validated in AI scenarios, SK Hynix is likely to sign large-scale long-term contracts with leading cloud providers.

Such long-term contracts can not only smooth price fluctuations but also lock in revenue, providing certainty for the company.

For investors, this means enhanced predictability of cash flow and profits, which will also increase the valuation multiples given by the market.

In other words, once the 321-layer QLC is successfully introduced into data centers, it will not only be a revenue growth point but also a factor in the reconstruction of the valuation system.

It is noteworthy that the storage industry must also face geopolitical uncertainties.

The policy game among the United States, Japan, South Korea, and China in the semiconductor field has made supply chain security an important issue.

SK Hynix's choice to be the first to break through 300 layers is not only a corporate competitive strategy but also has the shadow of industrial policy.

In the context of the increasing global emphasis on AI infrastructure, whoever can master the mass production capability of high-density NAND will occupy a more advantageous position in national strategic competition.

This logic of "technology as geopolitical leverage" will inevitably reflect in the risk premium levels of the capital market.

From a capital perspective, there are at least three investment logics worth noting regarding the 321-layer QLC NAND: first is the cyclical logic, where counter-cyclical technological breakthroughs during industry troughs are expected to catalyze a new upward cycle.

Secondly, of course, is the demand logic, where the growth certainty of AI data centers finds structural incremental markets for products; third is the pattern logic, where being the first to enter AI applications helps enhance SK Hynix's pricing power in the global NAND market The result of the overlap of these three logics indicates that this product launch is not merely a technical event, but carries deeper significance that attracts the attention of the capital market.

In the long evolution of the semiconductor industry, memory chips are often seen as the most brutal track.

This field reflects the challenges of process limits and has always been an amplifier of capital cycles.

SK Hynix's launch of 321-layer QLC NAND is a gamble at the intersection of technology, market, and capital logic.

If the demand from AI data centers explodes as expected, this product will become a key support point for its profit recovery and market expansion; if demand falls short, the 321-layer QLC may become a symbol of over-investment risk.

For investors, this is both an opportunity and a typical uncertainty in the memory industry.

The famous British economist John Maynard Keynes once said in response to an investor's question in 1931: "The market can remain irrational longer than you can remain solvent."

In the memory industry, the art of investment has never been about predicting prices, but about identifying trend inflection points. SK Hynix's 321-layer QLC NAND is likely to be right at such a critical inflection point