The usual rule says a low P/E means a stock is cheap. In memory chips the rule tends to run backwards. A memory maker's P/E is often at its lowest right when earnings are at a peak, which is the worst time to buy, and at its highest or missing when earnings have collapsed, which has often been the best time. Micron's own history shows it twice in the last eight years. With memory makers now trading at 4 to 7 times forward earnings, it is worth knowing how that history went.
If you only read one thing
The short version: In August 2018, Micron's trailing P/E was 5.2, the lowest of its previous five years. The stock fell 34% over the next ten months. In June 2023, its P/E sat at the 98th percentile of its five-year range because earnings had collapsed. The stock rose 75% over the next eleven months. In memory, P/E measures where earnings are in the cycle more than whether the stock is cheap.
Why memory breaks the P/E ratio
P/E divides the price by the last twelve months of earnings. That works when earnings are fairly stable. Memory earnings are not. DRAM and NAND are sold in large part at prices that reset every quarter, so when supply runs short, prices and margins jump together, and when new factories come online, both collapse.
Micron's last full swing shows the size of it. In fiscal 2022 it had revenue of $30.8 billion. In fiscal 2023, revenue fell to $15.5 billion and Micron lost $5.8 billion. A P/E built on fiscal 2022 earnings described a business that no longer existed a year later.
The takeaway: Memory earnings can halve, or flip to losses, within a year. A P/E built on peak earnings makes the stock look cheapest exactly when those earnings are about to fall.
Micron's own history
We rank each week's multiple against the trailing five years of weekly values, using only data available at that date. Here is Micron at four points:
| Week | Price | P/E (5y %ile) | EV/Sales (5y %ile) | What came next |
|---|---|---|---|---|
| Aug 6, 2018 | $51.37 | 5.2 (0th) | 2.3 (20th) | Fell 34% by June 2019 |
| Aug 1, 2022 | $62.46 | 7.1 (35th) | 2.1 (26th) | Revenue halved in fiscal 2023 |
| Jun 5, 2023 | $65.43 | 46.4 (98th) | 3.2 (77th) | Rose 75% by April 2024 |
| Oct 5, 2026 | $1,074.89 | 24.3 (59th)* | 13.4 (96th) | ? |
The 2018 row is the textbook case. Micron's P/E was the lowest it had been in five years, near the 2018 memory-price peak. The stock fell about a third as DRAM prices dropped. The 2023 row is the mirror image: earnings had nearly disappeared, so the P/E was at the top of its range, and that turned out to be close to the bottom for the stock.
Seagate shows the same pattern with hard drives. In March 2022 its P/E was 11.3, in the lower half of its five-year range. By late October 2022 the stock had fallen from $89 to $51, and by mid-2023 Seagate was losing money and had no P/E at all. You can check both on Micron's P/E history and Seagate's P/E history.
The takeaway: Micron's lowest P/E in five years came near the 2018 peak, and its highest came near the 2023 bottom. Buying on a low P/E would have meant buying the top.
Sales multiples are steadier, not magic
This is why our AI memory theme page ranks on EV/Sales. Revenue swings less than earnings, EV/Sales still has a value when a company loses money, and it counts debt and cash. Our guide on when to use P/E, P/S, EV/Sales, or EV/EBITDA covers the general case.
It does not time the cycle either. Look at the 2018 row again: EV/Sales was at the 20th percentile, which also looked cheap, because revenue was inflated by peak prices too. No trailing multiple tells you where the cycle is. What a sales multiple does is keep a reading through the whole cycle, so you can compare today with the last peak and the last trough on the same scale.
On that scale, today is different from 2018. Then, investors paid little for each dollar of peak revenue. Now they pay more than at any time in Micron's last five years. That is the market betting the shortage lasts longer than past ones did.
The takeaway: EV/Sales is steadier than P/E for memory stocks, but it is not a timing tool. Its use is comparison: today Micron's EV/Sales is at the 96th percentile, the opposite of 2018's 20th.
What to check instead
- Contract prices. TrendForce publishes quarterly DRAM and NAND contract price forecasts. For the fourth quarter of 2026 it expects another 10% to 15% for ordinary DRAM and 15% to 20% for NAND. The first forecast of falling prices matters more than any multiple.
- Capacity plans. Cycles end when new factories ship. Micron's prepared remarks put its next Idaho fab's first output in mid-2027 and a second in late 2028.
- Forward versus trailing. When forward P/E sits far below trailing P/E, the market expects earnings to keep rising. Micron trades at about 5.7 times forward earnings, SK hynix and Samsung at about 3.9. Those low numbers carry the same warning as 2018: they assume record margins hold.
- Inventory. Rising inventory days at the memory makers usually come before price cuts. Micron said it expects its inventory days to keep declining.
The takeaway: For memory stocks, watch prices and supply before multiples. A cheap-looking P/E is often the last thing to change before a downturn.
Every memory name we track is on the AI memory theme page, ranked on EV/Sales with a P/E toggle so you can see both views side by side. The full valuation table is in our AI memory stocks valuation percentiles post. If band charts are new to you, read how to read a valuation band chart first.
Data: SEC EDGAR XBRL filings and market prices, computed by StockResearch. Historical percentiles use only the five years of weekly data available at each date. Prices are weekly closes. Jake is the founder of StockResearch.app, where he writes data-first research on valuation dislocations. Nothing here is investment advice or a recommendation to buy or sell any security. Do your own research.