Author: Wang Chuan
This article is a continuation of Wang Chuan: When the Neighbor Old Wang Made 30x on Memory Stocks, How to Avoid Anxiety (Part Five) - The Bullwhip Effect.
1/ Memory company stocks are things that can easily thrill inexperienced youngsters to climax. Over thirty years ago, the primary medium for mobile storage in the computer industry was the 3.5-inch, 1.44 MB floppy disk. By the end of 1994, a company named Iomega launched a 100 MB removable hard drive, also known as the Zip drive, priced at $199. The Zip drive solved a major problem for consumers who needed to back up and transfer large files.
2/ Iomega's sales skyrocketed from $140 million in 1994 to $1.21 billion in 1996. Its stock price also surged from around $2 per share at the end of 1994 to an equivalent of $330 per share (accounting for stock splits) by May 1996, delivering a return of over 160x in a year and a half. I recall a sincere sigh from a netizen on a tech stock forum back then: "This is even more beautiful than sex!"
3/ After May 1996, Iomega's stock declined continuously, falling over 85% from its 1996 peak by the end of 1999. Eventually, in 2008, EMC acquired Iomega for $210 million, representing a 97% drop from its peak market capitalization of $7 billion in 1996. What does a 97% drop mean? It means that after an 85% drop, some speculators think they're getting a bargain and jump in, only to suffer another 80% decline.
4/ The main theories held by staunch Iomega bulls back then were twofold: i) Until 1996, potential competitors seemed weak and expensive. ii) There was a genuine possibility that the Zip drive could become a standard PC component, like the floppy disk drive. If this could attract hundreds of millions of users, with each Zip disk generating over ten dollars in profit, the future would be limitless. Iomega's stock became the first meme stock of the internet era to attract massive retail investor frenzy. A flood of capital created a self-reinforcing trend, burying many short sellers.
5/ Iomega's downfall was more complex than outsiders imagined. The latter half of 1996 saw a simple price correction; competitors were barely visible. Revenue in 1997 was $1.74 billion, with growth already noticeably slowing. CD-R, as a potential competitor, had become cheaper than Zip disks for high-end users. By 1998, CD burner prices were close to Zip drives, but a CD cost less than a dollar, far cheaper than a Zip disk, completely destroying Iomega's moat. While its 1998 revenue only fell 3% from 1997, gross margin dropped from 31% to 25%, turning profit into loss. The story was over.
6/ Storage industry products, epitomized by Dynamic Random-Access Memory (DRAM), are among the most commoditized in the tech sector. Commoditization means no brand premium, with prices fluctuating rapidly based on global supply changes. Historically, DRAM chip prices have plummeted over 80% within short periods at least six times: 1985, 1998, 2001, 2009, 2012, 2023, with numerous other instances of 30% to 50% declines in between. The stock price declines of memory companies are even more brutal than chip price drops, with 95% crashes or bankruptcies being commonplace. Micron's stock price in May 2025 was roughly the same as in June 2000, losing a full 25 years. Four major bankruptcies in the storage industry over the past three decades include: i) Mostek, 1986. ii) Qimonda, 2009. iii) Spansion, 2009. iv) Elpida, 2012. Countless smaller companies have also gone bankrupt.
7/ The nature of the storage industry is elastic demand facing capital-intensive, long-cycle, rigid supply. When storage prices are too high, elastic demand naturally recedes, finding ways to circumvent it. However, when rigid supply comes online 18 months later, capacity must run at full tilt, and products must be sold immediately at any price to maximize profit. Once rigid supply slightly exceeds elastic demand, price drops occur immediately, sometimes violently.
8/ Looking back, the surge in the entire memory sector's stock prices starting in September 2025 was essentially because demand from cloud service providers for various memory types consumed by AI chips, especially High Bandwidth Memory (HBM), crossed a critical threshold. To lock in capacity for 2026 and 2027, cloud providers were willing to accept significant price hikes. As long as one or two buyers were sufficiently frantic, competitors were forced to follow suit. The panic over shortages quickly spread to smaller buyers and the consumer electronics sector. Learning from past price crashes, major memory makers didn't rapidly increase capacity. Instead, they opted to let prices skyrocket, exploiting the temporary advantage to make a killing while time was still on their side.
9/ In Q1 2026 (non-fiscal year), flash memory maker SanDisk's production cost was $1.288 billion, compared to $1.313 billion in the same period last year, meaning costs actually fell about 2% year-over-year, with the volume of storage produced remaining largely unchanged. However, Q1 2026 revenue was $5.95 billion with a gross margin of 78.3%, whereas the year-ago quarter saw revenue of $1.695 billion and a gross margin of only 22.5%. Therefore, the 251% revenue growth came primarily from corresponding storage price increases, not from selling more goods.
10/ Why did prices more than double without selling much more? Because market demand suddenly exploded, while supply was rigid, limited to what was available in the short term, leading to a scramble for limited supply that drove up prices. This implies a counterintuitive future phenomenon: When rigid supply finally catches up and rebalances with demand, flash memory prices and gross margins will inevitably return to previous levels. At that point, although sales volume increases, total sales revenue and net profit will actually decrease. The more you sell, the less you earn.
11/ Similarly, for Micron in the quarter from November 1, 2025, to the end of February 2026, operating costs were $6.1 billion, a less than 20% increase from $5.09 billion in the year-ago quarter. However, sales revenue was $23.86 billion, nearly triple the $8.05 billion from a year ago. Gross margin was 74.4%, compared to only 36.8% a year earlier. Micron's product line includes HBM, DRAM, and flash memory, making its capacity constraints and price dynamics more complex than SanDisk's, but it cannot escape the same underlying logic.
12/ For commoditized goods, high margins themselves destroy high margins, and high prices themselves erode marginal demand. Seeing gross margins above 70%, major memory makers could no longer remain passive. Starting in 2026, they invested tens of billions of dollars to increase capacity, though a significant portion of new capacity is expected to come online only in the second half of 2027.
13/ Those bullish on memory stocks might argue that memory companies have started signing long-term agreements with customers to lock in capacity prices, which should mitigate the risk of price collapse, right? The truth is, the less stable a relationship, the more eager parties are to sign long-term agreements. The reason for signing them is the temporary, intense need to avoid worst-case scenarios, creating a false yet fragile sense of security for both sides. However, when circumstances change substantially, the stronger party under the new conditions will generally find some excuse in the agreement and immediately renege. The so-called long-term agreements in the memory industry typically don't exceed five years. Once future memory capacity comes online and spot prices fall below long-term agreement prices, buyers can exploit various loopholes in the agreements to force memory makers to share the pain of the price drop immediately. Even if the contract is legally airtight, buyers can threaten to shift more business to competitors who offer better terms once the agreement expires. Memory companies will usually concede immediately to protect long-term interests. Long-term agreements signed between memory makers and customers are probably about as effective as the Molotov–Ribbentrop Pact (the 1939 Nazi-Soviet non-aggression treaty). When everyone senses risk and rushes to use formal agreements to hedge against it, don't naively think the risk is mitigated. This is precisely a signal that risk is increasing.
14/ There's also an asymmetry here: When all players are making huge profits, it only takes one new player indifferent to short-term economics and capable of sustaining long-term losses to enter the market and change the supply-demand dynamic; or one new technological innovation to emerge, drastically reducing demand. You cannot predict beforehand which specific factor will directly shift the balance. You just need to know that the risk of memory price declines relative to the possibility of further increases is now highly asymmetric. That's enough. Risk factors include but are not limited to: i) Economic recession due to rising interest rates and inflation. ii) Cloud service providers cutting capital expenditures on AI. iii) New memory capacity coming online faster than expected, especially from Chinese companies adept at massively ramping up capacity regardless of cost, like CMXT and YMTC. iv) New AI chip designs, model architectures, and software algorithms that can drastically reduce memory requirements. In fact, precisely because of soaring memory prices, smart minds worldwide are brainstorming at all levels—chip design, model architecture, software algorithms—to actively reduce memory demand.
15/ The memory industry, like other highly commoditized sectors, has another deadly trap: at the peak of the cycle, product profits are extremely high, but company P/E ratios are often very low, sometimes even single-digit. This appears to be a good value investment, but it's actually the moment of greatest risk. Because once commodity prices crash, previous profits quickly shrink or even turn into losses, rendering the low P/E ratio meaningless. There will always be simple-minded investors tempted by low P/E ratios who, without deeper thought, willingly plunge their entire savings into this colossal wealth incinerator during the industry's downturn.
Neighbor Old Wang is still lost in the dream of effortlessly striking it rich with memory stocks and exiting unscathed in the future. Don't disturb him. To be continued.
(To be continued)
All articles represent the author's personal views for reference only and do not constitute investment advice for the mentioned assets. Investing carries risks, and one must be cautious when entering the market.
