Recently, while researching the memory industry, I found that many people are unclear about which memory stocks exist and how to choose among them. In fact, this field may look complicated, but once you understand the logic of the industry chain, you can find targets that fit you.

Memory stocks can basically be divided into three categories. The first category is manufacturers that directly produce chips—Taiwanese companies such as Nanya Technology, Winbond, and Macronix, as well as global giants such as Micron, Samsung, and SK Hynix. These companies have the highest profit elasticity, but they are also the most volatile; business-cycle conditions have a deep impact on them. The second category is companies that control ICs and make modules—such as Phison, Adata, and Transcend. Their profits are relatively steadier, because they hold a moat in software integration. The third category is companies that focus on specialized areas—for example, Macronix, which has deeply specialized in NOR Flash and embedded storage; risks are relatively lower.

When it comes to the global landscape, Samsung, SK Hynix, and Micron almost dominate the entire market. According to the latest rankings, Samsung leads by a wide margin with a market cap of nearly $900 billion. SK Hynix follows closely, and Micron ranks third. Interestingly, HBM (High Bandwidth Memory), a key product in the AI era, is currently led by Hynix and Micron. When these big manufacturers put all their capacity into AI chips, Taiwanese companies actually have an opportunity to benefit from order-transfer dividends.

Why are memory stocks so volatile? The core reason is that this industry is always cyclical: shortage → capacity expansion → oversupply → price collapse → production cuts → back to shortage. This cycle runs through once every few years. In addition, building a wafer fab costs several tens of billions of dollars, and if you miss the investment timing slightly, when capacity comes online, the market may already have reversed. Most importantly, the global memory market is dominated by only a few companies, and their decisions directly determine the price cycle.

Now let’s get into which specific memory stocks are worth watching. For Taiwan stocks, Nanya Technology is the purest DRAM concept stock, focusing on DRAM manufacturing, with AI applications already becoming a major growth driver. Winbond follows a niche route: it focuses on special DRAM and NOR Flash, avoiding a price war in general-purpose DRAM. Phison has the highest NAND Flash purity; the current supply gap is still close to 20%, so it is difficult for the short-term supply-and-demand imbalance to change. Powerchip makes memory wafer foundry products; it has cooperation with Micron, but the overall competition in the semiconductor foundry market is intense. Macronix, meanwhile, focuses deeply on NOR Flash and ROM. It has technological advantages in the automotive and industrial sectors, so industry volatility is relatively lower.

In the U.S. market, Micron is the only U.S. company with large-scale DRAM and NAND manufacturing capacity. As HBM capacity continues to expand, profits are clearly recovering. SK Hynix is the global leader in DRAM and HBM shipments; HBM3e and HBM4 have already entered mass production first, directly benefiting from the explosion in AI compute demand. Kioxia, formerly Toshiba Memory, is the main NAND Flash supplier worldwide. In the past six months, its market cap jumped from 43rd globally to 10th. After Western Digital completed the split of its NAND business in 2025, it transformed into a pure HDD company focused on hard drives. Its 40TB enterprise-grade HDDs have already sold out for the year.

Memory stocks and AI stocks are actually two different things. Memory stocks are cyclical trading assets—you earn from the rhythm of the business cycle. AI stocks are trend investments—you earn from structural growth. The holding strategy for memory stocks is to gradually build positions at the bottom of the cycle, then step out gradually when prices surge sharply and market sentiment gets overheated. Currently, memory prices are still trending upward, and the tight supply situation on the supply side is unlikely to ease in the short term. Related targets tied to the manufacturing and module segments still have upside momentum.

If you want to judge when it is most worth entering, you can watch for several signals. First, when DRAM prices stop falling—this is the most important signal of a sector turning point. Second, when leading manufacturers begin cutting production, which indicates supply is shrinking. Third, when inventories decline: as inventory days fall from their peak, it means downstream demand is starting to warm up. Currently, inventories of global memory original equipment manufacturers are at historic lows; some major companies have only about 4 weeks of inventory left in total. This is why prices are prone to rise but difficult to fall.

In the end, memory stocks are not stable growth stocks, but cyclical trading assets. What you need to judge is not which company you should hold forever, but where the business cycle currently stands. In the last cycle, the memory stocks that fell deeply became big winners in this round because of AI supply gaps. The essence of memory stocks is to profit from the rhythm, not from the company itself. You can open a simulated account right now and, over the next few weeks, observe the price trends of DRAM contracts, track major manufacturers’ earnings reports and capital expenditure directions, and practice judging which phase the memory cycle is in. Once you understand the cycle more clearly, then consider using small capital to carry out real trading.