I’ve been keeping a close eye on the memory chip sector lately and found that many people actually don’t understand why the same kind of memory “concept stocks” can have such wildly different volatility. The real reason is that they play completely different roles in the supply chain.

The memory industry is broadly divided into three layers. The first layer is the chip manufacturers that directly produce chips—such as Nanya Technology and Winbond. As soon as market quotes rise, their profit sensitivity is the highest, but at the same time they also bear the greatest cyclical risk. The second layer is companies that control ICs and modules—such as Phison and Adata. These firms build moats by integrating software and establishing system protocols, so their profits are usually more stable. The third layer is the global giants: Micron, Samsung, and SK Hynix. These three companies monopolize more than 94% of the global DRAM market and control pricing power.

Personally, I think if you want to profit from price spreads, you should focus on chip manufacturers. If you want something steadier, choose module makers. And if you’re looking at long-term trends and trying to catch big moves, then the U.S. giants are the key.

Speaking of global rankings, as of the most recent figures, Samsung leads by a wide margin with a market cap of $897 billion, followed closely by SK Hynix and Micron. Interestingly, Kioxia, the memory company that was originally part of Toshiba, saw its market cap jump from 43rd place to 10th place within half a year. Behind that is the explosive demand for high-end SSDs from AI data centers. On the Taiwan side, Nanya Technology, Winbond, Macronix, and Powertech have all squeezed into the global top ten, showing that Taiwan truly has global competitiveness in the memory chip field.

Why do memory stocks swing so much? Put simply, the industry has a cycle that’s unavoidable: shortage → capacity expansion → oversupply → price collapse → production cuts → shortage again. The latest forecasts show that this year’s Q2 DRAM and NAND prices are expected to rise 51% and 50% respectively quarter over quarter—far higher than earlier predictions. On top of that, building a memory wafer fab can cost several tens of billions of dollars in many cases. Once you get the timing wrong, by the time production capacity ramps up, the market may have already reversed—so this industry is always betting on timing.

In the U.S., Micron is the largest memory chip manufacturer in the country. It produces both DRAM and NAND, making it a storage play with strong upside flexibility. HBM capacity continues to expand, and memory quotes are entering an upcycle; overall profitability is visibly recovering. SK Hynix is the global leader in HBM shipments. HBM3e and HBM4 have already moved into mass production, directly benefiting from the explosive surge in AI high-computing-power demand. Rambus Technology focuses on DDR5 and HBM memory buffer chips, holds a quasi-monopolistic position in that area, and has very clear growth momentum.

In Taiwan, Nanya Technology is one of the few companies that focuses on DRAM manufacturing. Its customized AI memory products have already started contributing to revenue. Winbond focuses on niche DRAM and NOR Flash, avoiding the price-cut competition common to general-purpose DRAM, so overall profit fluctuations are relatively smaller. Phison is one of Taiwan’s highest-purity NAND Flash companies. At present, the NAND supply shortfall is still close to 20%, and AI inference is driving data storage demand that is almost unlimited—so in the near term it’s still difficult to change the “tight supply and strong demand” situation.

Memory stocks and AI stocks are actually completely different things. Memory stocks are cyclical trading assets, while AI stocks are trend investments. Memory stocks profit from the rhythm of economic cycles; AI stocks profit from the upside of structural growth.

So when is the best time to buy? Usually, it’s when prices stop falling and the market is still pessimistic—then a rebound begins. You can watch for a few signals: whether spot DRAM prices have stopped falling and stabilized, whether leading manufacturers start cutting production, and whether days of inventory have fallen from their peak. Currently, global DRAM original-equipment inventory is at historic lows. Some major companies have only about 4 weeks of inventory left in total—this is a direct reason why prices are prone to rising but hard to fall.

To be honest, memory stocks aren’t stable growth stocks—they’re cyclical trading assets. What you need to judge is where the economic cycle is currently at, not which company you can hold forever. The memory-chip-related names that fell deeply in the previous cycle became a big surprise winner in this one because of AI-driven supply gaps. The essence of memory stocks is that you profit from the cycle’s timing, not from the company itself.

If you want to trade in swings, you can build positions slowly at the bottom of the cycle, then gradually exit as prices rise sharply and market sentiment gets overheated. Currently, memory prices are still rising, and the tight supply situation on the supply side is unlikely to ease in the short term. Holding related names on the manufacturing and module sides still has upside momentum. My suggestion is to first open a simulated account, use the next few weeks to observe the trend of DRAM contract prices, track the financial reports and capital expenditure directions of major memory manufacturers, and practice judging which stage of the memory cycle we’re in. Once you have a clearer grasp of the cycle, then consider real trading with a small amount of capital.