Coverage update · Memory · Micron FQ4 · Rating initiated
Micron’s Best Quarter Ever Wasn’t Enough
On September 28 we published a central test, alongside several supporting indicators: a fiscal-Q1 gross-margin guide at or above this quarter's roughly 86% keeps our thesis intact, and a guide below 80% breaks it.[1] Micron guided to 86.25% on a non-GAAP basis and 85.95% on a GAAP basis, alongside the largest quarter in its history: $54.23 billion of revenue, an 87.0% gross margin, and $33.20 billion of adjusted free cash flow.[2] The test passed. The thesis is intact.
The stock is down about 30% from its June high and has given up its trillion-dollar valuation.[6] That is not a contradiction, and this note is mostly about why. We initiate at hold with a $1,090 twelve-to-eighteen-month target, about 24% above Friday’s close of $878.84.
One-page summary
| Item | Status |
|---|---|
| Rating | Hold. Initiated with this note; coverage resumed September 28 after closing on July 30. |
| Target | $1,090, twelve to eighteen months. Reaffirmed from our August work rather than raised, for the reason in section 4. |
| Price | $878.84, October 16 close; $1,082.28 on September 25, the trading day before the preview[7] |
| Implied return | About +24% to the target; the stock is about 30% below its June 25 high of $1,255[6] |
| The published test | Passed. FQ1 gross-margin guide 86.25% non-GAAP, against our “at or above roughly 86%” line and our “below 80%” break[1][2] |
| FQ4 result | Revenue $54.23B; gross margin 87.0% non-GAAP, 86.8% GAAP; non-GAAP EPS $33.42; adjusted free cash flow $33.20B[2] |
| FQ1 guide | Revenue $61.5B ± $1.5B; gross margin about 86.25% non-GAAP; non-GAAP EPS $38.15 ± $1.00[2] |
| What changed | Not the reported results. China's CXMT, a patent case at the ITC, and a market that has decided memory pricing is near its peak[4][5] |
1. The grade, condition by condition
Alongside that central line the preview set a break level and two supporting lists, of what would move us up and what would move us down. Grading only the headline would be the flattering version, so here is all of it.
| As published, September 28 | Result | Verdict |
|---|---|---|
| FQ1 gross-margin guide at or above roughly 86% keeps the thesis intact | 86.25% non-GAAP (85.95% GAAP) | Passed |
| A guide below 80% breaks the thesis | Not close | Not triggered |
| Down: revenue below $50.0 billion | $54.23B reported; $61.5B guided | Not triggered |
| Up: gross-margin guide at or above 86% with DRAM prices still rising | Spot DRAM has gone sideways; contract momentum is easing[3] | Half met |
| Up: fiscal 2027 estimates moving higher | Consensus FY27 EPS was $159.49 before the print, against $148.73 ninety days earlier; brokers raised price targets afterwards, which is not the same as raising earnings estimates[1][4] | Not yet verified |
| Down: a sequential DRAM price decline | Not yet in the data. Spot flat, contract momentum easing, one broker still forecasting double-digit fourth-quarter gains[3][5] | Not triggered, watch |
| Down: a ruling against Micron in the Netlist case at the ITC | No ruling. The investigation was instituted September 23 and further complaints followed[5] | Open |
The part the grade conceals
Our test defined a pass at roughly 86% and a break below 80%, and said nothing about the six points in between. An 82% or 84% guide was entirely plausible, and it would have returned no verdict at all. That is the same drafting error we identified in public three weeks earlier, when an inflation test printed inside a gap we had left between its own two branches — and we wrote at the time that a test which cannot answer for a plausible outcome is not yet a test. We then wrote another one with a hole in the middle.
For the record, the range we should have published: at or above roughly 86% supports the view; between 80% and roughly 86% requires us to re-examine the earnings estimate and the reason for the compression, without abandoning the thesis; below 80% breaks it. The four tests in section 5 are written so that every outcome lands on one side of a line.
Chart 1 — The Line We Published, and Where the Guide Landed
The guide cleared the pass line. It also stepped down from the quarter it followed.
Honest reading: a pass, and a narrow one. On the GAAP basis the guide is 85.95%, fractionally under the same line. Either way it is the first guided sequential decline in gross margin of this cycle, which is the fact the market has fastened onto.
2. The quarter itself was the largest in the company's history
Revenue of $54.23 billion beat the company's own June guide of $50.0 billion by 8.5% and consensus of $51.24 billion by 5.8%. Non-GAAP earnings of $33.42 a share beat the $31.59 consensus. Operating cash flow was $43.97 billion and adjusted free cash flow $33.20 billion in a single quarter.[2] Guidance for the quarter now running calls for $61.5 billion of revenue, 13% above the quarter just reported, and $38.15 of earnings.
Chart 2 — Guide, Consensus, Actual
Micron beat its own guidance on every line that matters, then guided higher again.
On revenue and non-GAAP earnings the quarter beat both guidance and consensus. The guided step down in margin is the one number that gives the peak-profitability argument something to work with.
3. So why did the stock fall?
Three reasons, and we should be straight that our own August work named two of them as the bear case before any of this happened.
China. ChangXin Memory Technologies announced progress on a denser DRAM design and signalled server-memory products using it by the end of 2026. Citi has said mass production is unlikely before 2028 and Goldman Sachs has called the threat exaggerated, but the market did not wait for the argument to be settled.[5]
The patent case. Netlist's complaints at the International Trade Commission, covering high-bandwidth memory and DDR5 products, add a legal risk that could take years to resolve and carries material downside. There has been no ruling. The investigation was instituted on September 23.[5]
The cycle. The selloff is consistent with growing concern that memory pricing is at or near its peak, and the guided step down in gross margin gave that view its first number from the company itself. Spot DRAM has gone sideways and contract momentum is easing, even as one broker still forecasts double-digit fourth-quarter contract gains.[3][5] Rising rates and a tech tape that shed more than a trillion dollars in value did the rest.[5]
Chart 3 — Three Weeks, Two Ends of Our Own Range
On September 25 the stock sat at the ceiling of what we would defend. It now sits near the floor.
Nothing in the band moved. The price travelled across it in three weeks, from within a few dollars of the ceiling to within two dollars of the floor. The question has changed with it: not whether Micron deserves to trade above our ceiling, but whether the market is right that the floor should be lower.
4. The rating, and why hold
At $878.84 the stock trades on 5.51 times the fiscal-2027 consensus earnings that stood before the print. Our defended range for this business is 5.5 to 7.0 times next-year earnings, which puts fair value at $877 to $1,116 on those estimates. Friday’s close sits about two dollars above the floor of that range. Three weeks ago, on the eve of the print, it traded at $1,082.28 — inside the band, but in its top quarter. Our probability-weighted value is $1,090.
That leaves roughly 24% of upside, which is not obviously a hold. The reason we are not more constructive is the shape of the distribution rather than its midpoint: our own scenario work carries a bear case running to the $500s on a first cut to forward estimates, against a bull case we cannot underwrite until the supply agreements earn credit in a multiple. An upside of 24% against a downside of that size is not the asymmetry that earns a buy, and the two risks now pricing the stock — Chinese supply and the patent case — are precisely the ones that would widen the left tail rather than the right.
That combination argues for patience rather than enthusiasm, for one reason: the risks pricing the stock down are the risks our own work identified in August as the bear case, and they are not yet resolved. We wrote then that the whole question is whether peak earnings are a peak or a plateau, that the thesis breaks on the first cut to fiscal-2028 estimates rather than on any chart level, and that Chinese supply and the Netlist case were the two things that could do damage without a single customer cutting spending. Those are now the headlines. Being early to name a risk is not the same as knowing how it resolves.
Showing one piece of arithmetic we previously asserted. The preview described the valuation as the market “paying for about half of those earnings to last.” That is a claim about normalised earnings, and it deserves its working rather than a reader's trust. Normalising at twelve times mid-cycle earnings, our valuation range for this business across a full cycle, the September 25 price of $1,082.28 implied roughly $90 of permanent annual earnings, or 57% of the fiscal-2027 consensus of $159.49. At Friday’s $878.84 the same arithmetic implies about $73, or 46%. “About half” is the right order of magnitude; the precision belongs to the normalizing multiple, and ours is now on the page instead of behind the sentence.
On why the target is reaffirmed and not raised. A lower share price raises the upside to an unchanged target; it does not by itself raise fair value, and the estimate increase that would is the one item we have not verified since the results. So the target stays where it is until we hold a post-print consensus and can rework the assumptions behind it, rather than nudging a number on the strength of a quarter whose central metric has just stopped rising. The $1,090 is carried forward from work dated August 18, it has survived a round trip from $820 to $1,255 and back to $879, and it will be re-derived in full rather than nudged.
5. What would change this view
Four indicators we will track. Each is written so that any outcome falls on one side of a stated line, but the rating reflects the weight of the evidence together rather than a mechanical response to any single threshold.
| Test | Evidence that would strengthen the thesis | Evidence that would weaken it |
|---|---|---|
| Fiscal 2027 consensus EPS, at our next update | At or above $159.49, the level before the print: unchanged estimates would suggest the decline has been a lower multiple rather than lower near-term earnings expectations | Below $159.49: the first cut to the near-term estimate. This governs our near-term view; a cut to fiscal-2028 estimates would matter more, because that is what tests whether elevated earnings persist |
| FQ2 gross-margin guide, reported in December | At or above 84%: compression stays inside the range we are prepared to tolerate. The 84% line is our judgment, not a model output | Below 84%: two consecutive declines make it a trend and the multiple should compress further |
| The Netlist case at the ITC | An outcome short of an exclusion order leaves the thesis as written, though an adverse ruling of narrower scope would still warrant review | An exclusion order against Micron products triggers an immediate review regardless of price |
| CXMT | No verified server-DRAM product shipping in volume before the end of 2027 | Verified volume shipment inside that window changes the supply picture and the defended multiple with it |
6. What makes this note wrong
We hold Micron, and it is the largest position in the book we write about, so a hold rating on a stock we own and have not sold deserves more skepticism than a rating on something we do not. The target is reaffirmed from work done at a different price with different estimates, which is a defensible choice and also a convenient one. Our defended multiple is a judgment drawn from memory's history of not sustaining peak multiples, and if the long-term supply agreements genuinely broke that cycle, the range is too low and the Street's $1,555 average is nearer the mark than ours. And the test we published measured the company when the thing that moved the stock was the cycle; passing it told us less than we expected it to, which is worth remembering the next time we write one.
Sources
- bpleon research, “Micron reports Wednesday. The margin guide decides our view,” September 28, 2026 — the test wording, the FY2027 consensus of $159.49 against $148.73 ninety days earlier, and the October 2 commitment. Read the preview.
- Micron Technology, fiscal Q4 2026 results, September 30, 2026 (Form 8-K, exhibit 99.1): revenue $54.23B; GAAP gross margin 86.8%, non-GAAP 87.0%; GAAP diluted EPS $32.87, non-GAAP $33.42; operating cash flow $43.97B; adjusted free cash flow $33.20B. Fiscal Q1 2027 guidance, quoted verbatim: revenue “$61.5 billion ± $1.5 billion”; gross margin GAAP “Approximately 85.95%,” non-GAAP “Approximately 86.25%”; diluted EPS GAAP “$37.84 ± $1.00,” non-GAAP “$38.15 ± $1.00.” SEC filing.
- TrendForce memory spot and contract updates, October 2026: DRAM spot trading quiet and broadly flat; contract price momentum easing as multi-year server agreements move to band mechanisms.
- Post-print broker actions, September 30 to October 16, 2026: targets raised at several firms, including an upgrade to buy with a $1,250 target; published average target around $1,555 with a $1,550 median. Third-party estimates, which we have not independently verified.
- Coverage of the October decline, October 13 to 17, 2026: ChangXin Memory Technologies' denser DRAM design and end-2026 server-memory signalling; Citi's view that mass production is unlikely before 2028 and Goldman Sachs' view that the threat is exaggerated; Netlist complaints at the International Trade Commission covering HBM and DDR5, investigation instituted September 23; peak-pricing concern and a broad technology drawdown.
- Price history: high of $1,255.00 on June 25, 2026; the stock has since given up its trillion-dollar market value.
- Prices: $1,082.28 September 25 close; $878.84 October 16 close, down 5.63% on the day, against a 52-week range of $192.40 to $1,255.00.
Disclosures. This note is research and opinion, published to all readers alike. It is not investment advice for any person. The author holds a long position in Micron, held continuously since before our July 30 note. A position in a memory-sector ETF (DRAM) was also held with a call option written against it; a protective stop on that position was triggered on October 19 and the resulting exposure is being reconciled, so it should not be assumed unchanged at the date of this note. bpleon research received no compensation from Micron or from any party with an interest in it. Company guidance figures are Micron's own and are quoted from its filing; consensus and broker figures are third-party estimates and are labelled as such. Tests are graded against their published wording rather than their intent.