Recently, U.S. memory stocks have continued to slide: a range of storage names such as Micron, Western Digital, Seagate, and others remain under sustained pressure. Market sentiment has abruptly shifted from the frenzy a year ago over AI computing power to panic about an upcoming downturn in the storage cycle. Some fans have asked me: “Isn’t AI still booming? Aren’t storage companies the core beneficiaries under ‘compute power scarcity’?” Then why are the stock prices still falling relentlessly?
I’ve seen too many similar scenarios: when the narrative is at its loudest, risk is often at its highest; and when panic spreads and nobody’s paying attention, opportunities are actually beginning to take shape. In this round of declines in storage stocks, is the underlying picture already suffering irreversible deterioration, or is the cycle being forcefully shaken out? How should ordinary investors respond? If you’re free today, let’s chat about it.
Who is selling memory stocks? Three swords strike at the same time.
The decline in memory stocks isn’t caused by a single factor; it’s the combined result of three forces acting together.
First sword: traditional memory has oversupply; AI only rescued the “minority.”
The market treats “AI-driven memory demand” as one big story, but the reality is far harsher than the story. What AI truly boosts is HBM (high-bandwidth memory) and some server DDR5. But the biggest portion of the memory market—DRAM and NAND used in smartphones, PCs, and consumer electronics—has demand that remains weak.
In 2023, global memory market revenue plunged 37% year over year, and Micron’s net loss that fiscal year was as high as USD 5.83 billion. In 2024, while AI demand drove a surge in HBM prices, the recovery of traditional memory remains unconvincing. According to TrendForce/TrendForce Consulting, forecasts have been repeatedly lowered: consumer DRAM and NAND demand has stayed weak, while new capacity continues to be released. By the second half of 2024, contract prices for traditional memory had already started to loosen, and this trend has continued into the present.
In other words, the market is finally starting to realize: AI isn’t a savior for all memory companies. Funds have begun to shift from “buying across the board” to “carefully scrutinizing,” and those U.S. memory stocks with lower HBM exposure are hit first.
Second sword: capacity expansion rebounds on itself, and pressure on the supply side surges.
Memory is a typical “cycle + heavy-asset” industry. During the industry’s deep losses in 2023, manufacturers were forced to cut production. In 2024, the super profits brought by HBM sparked a new round of capital expenditure competition. Micron, SK Hynix, and Samsung all announced large-scale expansions—not only expanding HBM, but also increasing traditional capacity.
The problem is that the construction cycle for memory capacity usually takes 2 to 3 years, while market demand can change within 1 year. The new capacity added in 2024 to 2025 is gradually becoming real supply pressure. The balance of supply and demand tilts from “tight” toward “loose,” and prices naturally come under pressure. Stock prices are a game of expectations: once the market sees that supply will be concentrated and released in the coming few quarters, funds withdraw early.
Third sword: high-beta assets are sold off in a high-interest-rate environment.
U.S. 30-year Treasury yields remain near historical highs. The interest-rate environment is unfriendly to highly valued, high-volatility growth stocks and cyclical stocks. Memory stocks are a typical high-beta asset: in a bull market they rally harder than the index, and in a bear market they fall worse than the index. In phases when liquidity tightens and risk appetite declines, memory stocks become a major target for institutional de-risking.
These three swords fall together—together they form the core logic behind the continued downtrend in memory stocks.
The essence of memory: a strong cycle, not a growth myth.
To understand why memory stocks are falling, you must first tear off the label of “AI growth stocks.” Memory chips aren’t Apple, and they aren’t software ecosystems like NVIDIA. They’re closer to commodities—the price is determined by supply and demand, and the cycle is extremely volatile.
Over the past two decades, the memory industry has gone through many brutal cycles: during the 2008 financial crisis, DRAM prices collapsed, and multiple vendors were on the brink of bankruptcy; in 2012 to 2013, the global memory industry underwent another deep reshuffle; in 2016, 2019, and 2022 to 2023, each downcycle came with price cuts of roughly half, inventory blowouts, and massive losses.
In 2023, Micron’s full-year net loss was USD 5.83 billion—typical of the bottom of the cycle. In fiscal 2024, Micron’s revenue rebounded to USD 25.11 billion, but net profit was only USD 0.778 billion, and profitability is far from returning to normal historical levels. What does this show? It shows that after going through a round of “weak recovery,” the memory industry hasn’t entered a true growth pathway—it only crawled out of the trough, and then faced new supply pressure.
Why the current market is so纠结 is because the AI narrative leads some investors to mistakenly believe memory has “passed the cycle.” But HBM’s share of the overall memory market is still limited, and traditional memory remains the core driver of the industry’s fundamentals. As long as supply and demand for traditional memory has not improved meaningfully, an HBM story alone can’t support a sustained, highly overvalued valuation for the whole sector.
The decline in memory stocks is, at its core, the market moving from an “AI narrative premium” back to “cycle realities.”
For the ongoing downtrend in memory stocks, my view is: don’t fight the cycle—wait for data to confirm the turning point; position management is more important than directional judgment; don’t let your strategy be captured by narratives, and don’t be dominated by panic.
A feasible strategy for ordinary investors 👀
Based on the analysis above, what should ordinary investors do?
First: strictly control total position size and eliminate a gambler’s mindset.
Memory stocks are extremely volatile; it’s not unusual for a single stock to drop by more than 30% or even 50%. If ordinary investors go all-in on one position, they can easily be shaken out in the bottom area. Recommendation: keep a single memory stock position to no more than 5% to 10% of investable assets, and limit the total allocation to semiconductor/cyclical assets to within 30%. Keep enough cash so you can stay proactive in extreme market conditions.
Second, distinguish HBM leaders from traditional memory manufacturers—don’t treat them as the same.
Companies with high HBM share—such as Micron and SK Hynix—don’t follow the same logic as companies like Western Digital and Seagate, which rely more on traditional memory. If you believe in the long-term AI trend, focus on stocks with strong HBM competitiveness; if you want to bet on a memory-cycle reversal, you need to closely track supply-and-demand indicators for traditional memory. You can’t buy indiscriminately just because “they’re all memory.”
Third: give up on trying to time the bottom precisely, and use batch-based investing (dollar-cost averaging).
No one can accurately predict the lowest point of memory stocks. Even industry insiders can’t foresee when prices stop falling. A more realistic approach for ordinary investors is to set an acceptable price range or valuation range, build positions in batches, and stretch the time horizon to 6 to 12 months. For example, after the stock price pulls back by a certain percentage from the peak, buy once every additional 5% to 10% drop, rather than firing all bullets at once.
Fourth: closely watch leading indicators—use data instead of emotions.
There are several key indicators in the memory industry worth tracking continuously:
· Contract price changes month over month for DRAM/NAND (whether they have stopped falling and started to rebound);
· Original-equipment manufacturer inventory days (whether it has fallen back from high levels);
· Capital expenditure guidance (whether there are clear cutbacks);
· Shipment volumes of smartphones, PCs, and servers (whether demand is warming up).
When these signals appear together—when contract prices turn positive month over month, inventory drops back to healthy levels, and original manufacturers begin proactive production cuts—the certainty of a cycle bottom increases significantly. Before that, any “buy-the-dip” involves an element of gambling.
Fifth: make good use of tools and diversify risk.
Experienced investors can use options to reduce the cost of entry—for example, selling put options to collect premium, or building a bull spread. Ordinary investors are generally better off participating through a semiconductor ETF to diversify away the risk of any single stock blowing up. But note that many semiconductor ETFs have concentrated weights in AI chip names such as Nvidia and Broadcom, so the actual exposure to memory stocks may be less than you imagine. Before buying, be sure to check the portfolio composition.
Sixth: maintain cash flow and patiently wait for right-side confirmation.
The duration of a cycle bottom is often longer than most people expect. The memory downturn from 2022 to 2023 lasted more than a year, and the winter of 2008 to 2009 lasted nearly two years. Until bottom signals are clear, the best strategy isn’t to overweight and bet on a reversal—it’s to preserve cash flow and wait for right-side confirmation. Missing the absolute low isn’t the real regret; what’s truly regrettable is running out of bullets too early and being forced out before dawn.
Seventh: the discipline learned from Web3 trading—stop-loss and take-profit are equally important.
Crypto taught me one harsh lesson: faith without a stop-loss ultimately often turns into losses you can’t bear. Memory stocks are the same. Before you buy, set up a stop-loss logic. If the fundamental thesis is disproven, exit decisively. After you’re profitable, take profits in batches—don’t hold indefinitely just because “this time is different.”
It’s not scary for memory stocks to fall. What’s scary is having no view of the cycle.
U.S. memory stocks are continuing to slide. On the surface, it’s the market adjusting the AI-memory narrative; on a deeper level, it’s the cycle reasserting control over pricing. Since the memory industry was born, it has never escaped the loop of: shortage →疯狂扩产 (furious capacity expansion) → oversupply → price collapse → capacity clearing. The explosion of AI only overlays a new demand variable on top of this ancient cycle, but it hasn’t changed the cycle’s nature.
For ordinary investors, what you most need right now isn’t predicting whether tomorrow’s price will rise or fall, but building your own framework for cycle analysis and trading discipline. Treat the cycle as a friend, data as your compass, and position management as the rule of survival. Falling is never the problem; the problem is whether you prepared before the decline.
The winter for memory stocks may still last for a while. But for investors who understand the cycle, stay patient, and adhere to discipline, every winter is the best seeding period before spring arrives.
PS: This article is based on publicly available information and industry logic and does not constitute any investment advice. DYOR~#美国存储股延续跌势