Memory prices plunge, memory stocks fall across the board—is the memory super cycle peaking?

Recently, the memory prices that had been rising for months suddenly turned downward, triggering concerns in the market about the memory cycle peaking.

According to market tracking data, several US retailers have seen large-scale price reductions in DDR5 memory, with single sets dropping by as much as $100. Taking Corsair’s VENGEANCE series as an example, the 32GB model with up to 6400MHz frequency is currently priced at about $379.99, dropping significantly from the recent peak of $490 and decreasing by more than $110 per set.

The domestic market has also been impacted. A wholesaler told China Business Journal that mainstream 16GB memory stick prices “dropped by more than 100 yuan in a single day,” and large traders who stockpiled earlier are now frantically selling off their holdings.

“Since last Saturday, prices have simply collapsed.” Mr. Wang, a wholesaler who has operated storage devices in the Buynow Market for years, told the media frankly. He showed an extreme price curve of a mainstream 16GB 3200MHz memory stick: last May it was only a little over 130 yuan, then surged all the way to a peak of 980 yuan in December. After months of fluctuations at the high, the current spot price has fallen back to around 700 yuan.

Mr. Wang lamented that since the price increases have exhausted consumer expectations, “If it’s not a necessity, people won’t buy. Compared to before November last year, our sales have dropped by more than 60%.”

Meanwhile, Google published a paper about a new compression algorithm called “TurboQuant.” The study noted that the technology can reduce the key-value cache (KV Cache) memory usage in large language models by at least 60%. Investors quickly priced in that: the AI hardware shortage problem will be fundamentally alleviated, and memory demand will be greatly reduced.

The chill in the spot market quickly spread to the capital markets. Micron Technology’s stock price has dropped more than 24% from its recent high, and Western Digital fell nearly 21% from a peak of $777.60. Last week alone, the US memory chip sector saw its market capitalization evaporate by nearly $100 billion.

Faced with plummeting prices and sharp declines in stock prices, market participants are seriously divided about the outlook of the memory industry. Some investors believe that the traditional memory “pig cycle” has peaked, while HSBC thinks worries are excessive, stating that we are currently in the middle of an AI-driven memory super-cycle, with strong demand for HBM and other high-end products, and memory shortages may continue for one to two years.

Buyers say “No”: Is the traditional “pig cycle” repeating?

For traders who follow the traditional cycle, the market slump isn’t that simple. Former journalist and noted semiconductor analyst in Taiwan, Dan Nystedt, pointed out that many bulls attribute the recent crash to Google’s paper, but that’s just superficial. Dan believes the real cause is that some smartphone memory chip prices stopped rising.

“The real reason is much simpler: some smartphone memory chip prices stopped rising. Buyers finally said ‘No’—this is the first sign of a peak that experienced memory cycle investors look for before selling.”

Dan Nystedt said. Due to high DRAM and NAND prices, some smartphone manufacturers plan to reduce or even discontinue production of mid and low-end phones in 2026. He revealed that buyers refused higher DDR4 prices two weeks ago.

Dan Nystedt likens the memory industry to the “pig cycle” in agriculture: High prices prompt companies to expand production, but building factories takes time, and when new capacity is released simultaneously, prices collapse. He believes investors following this script have quickly exited, causing Micron and SanDisk stock prices to retreat sharply.

Over the past 50 years, memory chips have undergone a dozen major boom/bust cycles. Since 2010, there have already been three: the 3G/4G and cloud computing boom from 2012-2015; the expansion of 5G and cloud service providers from 2016-2019; and the surge in PC/server demand during the pandemic from 2020-2023. The up-cycle starting in 2024 is driven by AI servers (HBM and SRAM).

“Whenever someone writes ‘this time is different’, that's usually a classic sign that bullish sentiment has become irrational.” Nystedt cited legendary trader Jesse Livermore’s famous saying: “The market is always right, but opinions are often wrong.” He cautioned investors that when chip buyers stop panic buying and rebounds repeatedly encounter sustained selling, smart money will swiftly exit in line with the script.

Structural Change: Memory companies are no longer “cyclical stocks”?

However, independent analyst Jukan disagrees with Dan Nystedt’s analysis.

He points out that buyer resistance to price hikes is mainly focused on traditional memory like DDR4, not the entire memory market. The abnormal spike in DDR4 prices was partly due to stockpiling in the Chinese market, which gave smartphone manufacturers room to adjust specs for low-end devices.

“But DDR5 is a completely different story.” Jukan said. Smartphone and PC makers obediently accepted large DDR5 price hikes in Q1 and even Q2 this year. In the current AI and high-end device ecosystem, DDR5 is not a bargaining chip for negotiation, but a core input buyers must secure—even at a premium. Flagship products built around DDR5 simply can’t reduce specifications.

Secondly, the market has completely overlooked the fundamental transformation in the business model of memory giants. Jukan scoffs at so-called “savvy investors” who blindly sell whenever spot prices drop.

“The operating mode of memory companies is no longer the blind expansion of the past.” Jukan keenly points out that the three giants—Samsung, SK Hynix, and Micron—are moving towards TSMC’s business model—that is, only after securing advance payments from key clients and long-term demand visibility, do they build new capacity.

Recently, Korean media reported that Samsung is discussing advance payment-based cooperation agreements with giants like Microsoft. Memory giants know better than anyone the pain of overcapacity destroying cycles. So now, they pursue highly restrained capacity expansion, rather than reflexive overbuilding.

Investment banks support: The memory super-cycle is just at halftime, the market’s five worries are overblown

Unlike the panic in the spot market, investment banks remain highly confident about the memory industry’s long-term prospects. HSBC’s research report released on March 30 explicitly stated, “In our view, the current worries are overblown; we are at the midpoint of an AI-driven super-cycle.”

The concerns in the current market are all overreactions, and the bank lists five specific worries:

1) Negative impacts from raw material and electricity price increases due to Middle East conflicts;

2) Slowing memory price growth in the second half of 2026;

3) Industry technologies like Google’s “TurboQuant” and Nvidia’s “KVTC” that reduce AI system memory usage;

4) Gradually increasing capital expenditure plans by major memory manufacturers;

5) Intensified competition from Chinese memory manufacturers.

The report says that Middle East conflict has no material impact on memory makers’ procurement of raw materials. The impact of absolute profit growth on stock prices is much greater than the slowing rate of DRAM price increases. At the same time, memory manufacturers remain highly prudent and restrained in executing capital spending.

Regarding Google’s TurboQuant technology, which triggered the market sell-off, the bank believes it’s too early for concern. The technology’s commercialization will take about a year, and its reference parameters are smaller than the current AI environment. More importantly, the bank points out, TurboQuant alleviates bandwidth bottlenecks, will improve system efficiency, and reduce token costs—thus accelerating commercialization and adoption of AI. The report states:

“The net impact is, we believe improved efficiency will accelerate AI development—this is a positive event and should trigger a sharp rise in AI adoption rate.”

At the same time, the bank projects that AI server shipments in 2026 will surge by 28% year-over-year. Between 2026 and 2027, the average DRAM content per server will grow by a robust 17%. As AI inference demand surges, enterprise SSDs (eSSD) are entering a golden age. The report forecasts that by 2027, eSSD will account for 40% of total NAND demand, up from 18% in 2023, with AI servers consuming 62% of that.

The bank believes that the current market is at the midpoint of an AI-driven super-cycle, comparable in scale to the six-year DRAM shortage caused by office automation from 1990-1995. In retrospect, from 1990 to 1995, with the popularization of Windows 3.0 and subsequent operating systems, office automation triggered a six-year structural shortage in DRAM, propelling the market size from $7 billion in 1990 to a sixfold surge of $41 billion in 1995.

Now, the bank believes that the memory shortage driven by infrastructure construction triggered by large models, Agentic AI, and physical AI (such as autonomous driving) will last at least one to two years.

Based on these judgments, the report is steadfastly optimistic about their benefit certainty in the memory super-cycle. As for the recent slump, the report writes: “We believe any correction is an additional buying opportunity.”

Risk warnings and disclaimersThe market has risks, and investment should be cautious. This article does not constitute personal investment advice, nor does it consider the special investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their specific circumstances. Any investment based on this is at your own risk.