Five numbers to start with: · Last quarter's revenue: $41.4B, +346% YoY, more than Micron's entire FY2024. · Consolidated gross margin: 84.6% — never seen in memory history (historical peak ~55–60%). · Quarterly EPS: $24.67. Next quarter revenue guidance: ~$50B (vs $11.3B a year ago). · Stock up 700% in one year, market cap above $1 trillion. · TTM P/E ~24x; annualized peak earnings push it to just over 10x — and that's exactly the issue.
Micron just reported a quarter whose numbers are almost too wild to be real: quarterly revenue $41.4B, up 346% YoY; consolidated gross margin 84.6%; EPS $24.67 – all records. The stock has 7x’d in a year, market cap above $1 trillion. As the only pure-play memory leader listed in the U.S., it has become the clearest proxy for the AI storage trade.
The most seductive line in the market: after all this rally, the P/E is still only around a dozen – isn't that cheap? That is exactly the classic value trap for memory stocks. To answer it, you first need to recognize where Micron sits in its cycle.
First, recognize the position: this is the top, not the bottom
Memory is the quintessential strong-cycle industry. At the trough, Micron's consolidated gross margin goes negative – in FY2023 it hit −11% with operating losses; at historical peaks it was around 55–60%. This quarter's 84.6% is a level never seen in this company or in all of memory history. Even more extreme: its single-quarter revenue of $41.4B already exceeds the full FY2024 total (about $25B).
The driver is AI. (high-bandwidth memory) is mandatory for AI accelerators; demand has exploded, supply is tight, pushing prices and unit value to extremes: HBM3E is sold out through 2026 and beyond; HBM4 will ramp starting March this year for Nvidia's Vera Rubin platform, with a ramp speed about twice that of the previous generation and yields beating expectations; on June 22, Micron signed a multi-year strategic supply-plus-investment agreement with Anthropic. The story is real, and it is strong.
Key cognitive difference: But remember one thing: an 84.6% gross margin is not the new normal; it's the hallmark of a cycle peak. It embeds the double amplification of an AI-HBM super-cycle plus supply tightness. Historically, every such peak has been followed by a mean reversion. We are closer to the "top" than the "bottom."
The illusion of a low P/E
Now back to that "cheap" argument. Micron's TTM P/E is about 24x; using annualized peak earnings, it's barely above 10x. That does look inexpensive. But the denominator of this P/E is the extreme peak earnings in memory history.
The script for memory cycles has been the same every time: trough production cuts → prices bottom → demand recovers → prices and gross margins surge → high prices trigger three oligopolists to expand capacity → oversupply → collapse again. Once the cycle turns down from the top, EPS can collapse by multiples, and a static P/E that looks low today can instantly become a very high P/E. So for cyclical stocks, using peak earnings to argue "cheap" via a low P/E is almost guaranteed to mislead. The correct anchor is normalized mid-cycle earnings.
The only variable that could rewrite the script: HBM
The bulls are not unaware of the cycle. They are betting this time is different – AI turns memory from a "commodity" into "part of compute." HBM capacity build cycles are long, validation barriers high, and deep binding to specific AI platforms with long-term contracts make supply release much slower than for ordinary DRAM, and prices stickier. If that holds, memory's mid-cycle profit margins could be permanently lifted a notch, and today's price might not look expensive against future normalized earnings.
The bears' counter-argument is equally powerful: no amount of demand can erase memory's destiny. An 84% gross margin is the strongest expansion signal there is – the three oligopolists cannot resist adding capacity when they see such margins. Once supply comes online, prices and gross margins will revert just like every past cycle. This binary debate – structural re-rating vs another cycle – determines what Micron is really worth.
So what is it worth?
For a cyclical stock, never use peak EPS times a P/E (that would yield absurdly high targets). Instead, use next fiscal year's normalized EPS times a cyclical P/E in three scenarios:
| Scenario | Weight | Key Assumptions | Target Price | vs Current |
|---|---|---|---|---|
| Bull | 30% | HBM re-rating, super-cycle continues | $1,350 | +27% |
| Base | 40% | High plateau then mild decline | $952 | −11% |
| Bear | 30% | Severe mean reversion in memory | $500 | −53% |
| Weighted | — | Probability-weighted expected value | $935 | −12% |
The three scenarios range from $500 to $1,350 – nearly a 3x span. This is not a "cheap or expensive" question; it's a binary bet on "structural HBM re-rating vs cyclical mean reversion." The weighted center is about $935, below the current price. In other words, the current price has already priced in a good chunk of the bullish case of "super-cycle continues + HBM re-rating," leaving very thin margin of safety for bulls.
My view: Micron is an impeccable company standing at an unsustainable peak. It is the core of AI storage and the only pure-play memory leader in the U.S.; the HBM story is real and strong. But an 84% gross margin and a 7x rally are the unmistakable hallmarks of a cycle top. The low P/E is a feature of peak earnings, not a sign of cheapness. I buy the story, but on price I'd rather wait: until the cycle offers clearer odds, I'm not chasing. The one question to watch: after the cycle corrects, can gross margin hold above 60%? If yes, the re-rating is real; if not, the 84% was just another transient peak.
If you're looking at this stock
Depending on your position, the approach differs (thoughts below are reference, not advice).
No position: Already +700%, market cap > $1T, at an extreme cycle top – no need to chase at current prices. Wait for cycle cooling or a clear pullback to build a small position gradually, capping it at 3–5% of portfolio.
Have a position, low cost basis: Enjoy the super-cycle, but set a reduction discipline – watch memory prices and gross margin; once prices turn down or inventories rise, start cashing out in tranches. Use covered calls / OTM puts to hedge the high volatility; don't try to sell at the exact top.
Want to short: The logic is valid but timing is extremely hard – the super-cycle and AI demand are still alive, HBM is sold out, so fighting the trend is costly. To express caution, options spreads are safer than naked shorts.
Data Sources & Disclaimer
Primary sources: Micron FQ3 FY2026 earnings (as of 2026-05-28, company IR / SEC 8-K) – revenue $41.46B (+346%), EPS $24.67, gross margin 84.6%, operating margin 80.4%; FQ4 guidance ~$50B revenue; HBM4 for NVIDIA Vera Rubin, HBM3E sold out through 2026+; Anthropic strategic supply + investment agreement (2026-06-22); stock price $1,064, market cap ~$1.22 trillion, +700% over the past year, as of 2026-07-01.
This article is a personal research note, published 2026-07-01, data as of 2026-07-01, and does not constitute investment advice. Normalized EPS, three-scenario target prices, cyclical P/E etc. are estimates or model-derived figures based on public data. Important note: the 84.6% gross margin and $41.5B quarterly revenue in FQ3'26 are extreme peaks from the AI-HBM super-cycle, far above historical memory peaks, and must not be extrapolated linearly. Micron is a high-volatility cyclical stock; judgment may become obsolete quickly as memory prices and new information change. Please make independent decisions. Market risk exists, invest prudently.
专注投资分析、市场洞察与资产配置。不追短期波动,只理解真正驱动长期回报的东西。
