The Semiconductor Cycle — Why Memory Prices Move Micron, Samsung and SK hynix
Key takeaways
- Memory chips (DRAM and NAND) are standardized products, so their prices swing widely with supply and demand, and these prices drive the profits of Micron, Samsung Electronics and SK hynix.
- The chip industry tends to repeat the pattern 'rising demand → rising prices → more investment → oversupply → falling prices,' which is why it is called a 'cycle.'
- Many analysts say HBM for AI servers, which is hard to make and sold largely under long-term contracts, is creating a pattern different from past cycles.
1. Two branches of semiconductors
| Category | What it does | Major companies |
|---|---|---|
| Memory | Stores data. DRAM (fast temporary storage), NAND (long-term storage, SSDs and smartphones) | Micron, Samsung Electronics, SK hynix |
| Non-memory (system semiconductors) | Computation and control. CPUs, GPUs, AI accelerators, smartphone APs, etc. | Nvidia, AMD, Intel, Qualcomm |
| Foundry | Manufactures chips designed by other companies | TSMC, Samsung Electronics (foundry division) |
| Equipment and materials | Equipment and materials used to make chips | ASML, Applied Materials, Lam Research, KLA, Tokyo Electron |
Note that Micron, Samsung Electronics and SK hynix earn most of their chip profits from memory, so they are the most exposed to this cycle.
2. Why the memory cycle happens
Memory is a standardized product whose specifications are nearly the same across companies. So its price is set by supply and demand.
- Rising demand: Sales of smartphones, PCs and servers increase, or new demand (AI) emerges.
- Rising prices and profits: Supply runs short, prices rise, and memory makers’ profits jump.
- More investment: Companies increase investment in fabs and equipment. A new fab takes one to two years or more to build.
- Oversupply: The added capacity comes online all at once, and when demand cools, inventory builds up.
- Falling prices and production cuts: Prices drop, profits shrink, and companies cut output. As inventory falls, the cycle returns to step 1.
The stock market usually tries to price in this pattern several months ahead. That is why share prices sometimes have already fallen when profits are at their best, or rise first while companies are still losing money.
3. What AI and HBM have changed
HBM (high-bandwidth memory) stacks multiple layers of DRAM to move data very quickly, and is used alongside AI accelerators (such as GPUs).
- It is hard to make, and yield (the share of chips produced without defects) is tricky to manage, so few companies can supply it.
- A relatively large share is sold under long-term contracts in which volumes and prices are set with customers in advance.
- Each wafer yields less HBM than ordinary DRAM, so increasing HBM output also has the effect of tightening the supply of ordinary DRAM.
For these reasons, there is both the view that “as long as AI demand continues, there may not be a sharp cycle like in the past” and the view that “if the pace of AI investment slows, the cycle will eventually return.” In October 2026, too, worries over big tech companies’ AI spending caused large swings in chip stocks.
4. Indicators investors watch
| Indicator | What it tells you |
|---|---|
| DRAM and NAND prices | The most direct signal of memory industry conditions. Contract and spot prices from market research firms |
| Korea’s chip exports | An early monthly read on global memory demand, released by Korea’s trade ministry around the 1st of each month. Check monthly figures on this site’s Semiconductor Cycle page |
| Philadelphia Semiconductor Index (SOX) | An index of the 30 largest semiconductor stocks listed in the U.S. (including foreign firms such as ASML and TSMC). Also sets the tone for Asian chip stocks the next morning |
| Big tech capital expenditure (capex) outlook | How much Microsoft, Google, Amazon, Meta and others spend on data centers → demand for AI chips |
| Inventory levels | Whether inventories at memory makers and their customers are building up or falling |
| Micron earnings | Reports weeks before most memory peers, so its results and outlook are an early read on the industry |
5. When looking at semiconductor ETFs
Semiconductor ETFs (for example, ones tracking the SOX index) are a way to spread an investment across many stocks. However, leveraged (2x, 3x) ETFs are designed to track a multiple of the daily return, so when prices rise and fall repeatedly, their performance can be much worse than the underlying index. For details, see “Leveraged ETF” in the Glossary.
This article explains the structure of the semiconductor industry. It is not a recommendation to buy or sell any particular stock.
All content on this site is for informational purposes only and is not investment advice. You are solely responsible for your investment decisions. Past performance does not guarantee future results. Editorial & Corrections Policy