Recently, I’ve been researching memory in particular, and I found that it’s definitely worth mastering. You know, even though they’re both semiconductors, the way to play memory stocks is completely different.

The memory industry chain is actually divided into three tiers. The top tier is made up of companies that directly produce chips—such as Nanya, Winbond, and Macronix. They have the greatest profit leverage, but they’re also the ones most easily crushed by the business cycle. The middle tier makes control chips and modules; companies like Phison and Transcend are comparatively more stable. The top tier is the global oligopolists—Micron, Samsung, and SK hynix hold the pricing power. Especially in HBM technology, Hynix and Micron are leading by a significant margin.

What’s most interesting is the ranking of global memory companies. As of this year’s April, Samsung’s market value is nearly $900 billion, far ahead, while SK hynix and Micron trail closely behind. But if you look at Taiwan stocks, Nanya, Winbond, and Macronix all make it into the global top ten, which shows that Taiwan factories really do have strong capabilities in niche products.

Why are memory stocks so volatile? In plain terms, it’s a dead loop: shortage → everyone rushes to expand production → oversupply → prices collapse and plunge → producers begin cutting production → then another shortage. This cycle turns once every few years. A recent Nomura forecast shows that in Q2 2026, DRAM and NAND prices will rise by 51% and 50%, respectively—much more than previously predicted. And globally, 94% of the DRAM market is monopolized by Samsung, Hynix, and Micron, while the five major companies control more than 80% of NAND Flash. With that kind of concentration, the decisions of a few companies can determine the entire price cycle.

In the U.S. stock market, I’m more bullish on Micron. It’s the only U.S. company that has large-scale DRAM and NAND production capacity at the same time. Its HBM capacity continues to expand, and overall profitability is clearly in a recovery phase. Hynix is even more aggressive—HBM3e and HBM4 are already in mass production, letting it directly capture the windfall from the AI compute capacity boom. There’s also a more niche company, Montage Technology, which specializes in buffer chips for DDR5 and HBM. To some extent, it’s a near-monopoly position, and its growth momentum is quite clear.

In Taiwan stocks, Nanya is the purest DRAM concept stock, and its AI memory products have already started contributing to revenue. Winbond takes a niche-focused route, avoiding the price-cutting competition of general-purpose DRAM, so its overall gross margin fluctuations are actually smaller. Phison is the highest-purity NAND player; the current supply gap is still close to 20%. The data storage demand brought by AI inference is almost limitless, so in the short term it’s difficult for the “supply-demand imbalance” pattern to change. Macronix has deep expertise in NOR Flash and automotive applications. This kind of embedded storage demand is relatively stable, which can be used to help balance the ups and downs of memory-cycle conditions.

To be honest, memory stocks aren’t stable growth stocks—they’re cyclical trading assets. What you need to judge is which stage the business cycle is currently in. The current situation is that DRAM prices are still trending upward; short-term supply tightness is hard to ease. And for those holding manufacturing and module-side targets, there is still upside momentum.

If you want to enter the market, I recommend focusing on three things: when will the DRAM contract price stop falling; whether the leading manufacturers have started cutting production (Samsung, Hynix, and Micron may have seen major earnings surges this year, but they’re all applying the brakes); and when will inventory days fall from their highs. At the moment, global factory inventories are at historical lows. Some major companies have only about 4 weeks of inventory left—this is why prices are easy to rise but hard to fall.

If you ask me, the profit from memory stocks comes from the timing, not from the company itself. Memory stocks that fell deeply in the previous round became big winners in this round because of the AI supply gap. You can start observing now: track DRAM contract prices and the quarterly reports of the main manufacturers over the next few weeks, and once you understand things more clearly, then consider making real moves.