Built by Grok 4.6 HighLam Research reported fiscal Q1 FY2027 after the close on Wednesday, October 21, 2026. Analysis prepared Monday, October 26, 2026. Quarter ended September 27, 2026. Fiscal year ends the last Sunday in June. This print is FY2027Q1, which is calendar 2026Q3. It is not calendar 2026Q4.
LRCX | Earnings Review — FQ1 FY2027
Verdict: ACCELERATING — the incremental dollar is memory systems, not China, and the street already owns the constrained next quarter. September revenue of $8,219 million grew +54.4% versus $5,324 million. That is the third straight step up in the year-over-year rate, and the latest step is +2,439 bps versus June’s +30.0%. Non-GAAP diluted EPS of $2.20 grew +74.6% versus $1.26. EPS did not turn with revenue. It kept fading into June (+36.8%) and only inflected this quarter, +3,776 bps. Non-GAAP operating margin of 40.2% is the first print above 40%, +520 bp versus 35.0%.
The beat is real and it is the smallest EPS surprise in the 12-quarter window. Versus FMP, revenue was +0.95% and non-GAAP EPS +$0.02 / +0.92%. The record is still 12-for-12 on both lines. Revenue surprises have settled near +1%. The EPS surprise fell from about +8% on each of the prior three prints because gross margin cleared a 52% guide by only 40 bp while the non-GAAP tax rate stepped up to 16.5%. GAAP diluted EPS of $2.19 is a penny under non-GAAP. This is not an add-back quarter.
What changed in the mix. Systems revenue of $5,733 million grew +61.6%. Customer support of $2,487 million grew +40.0% and did not give the growth back. Memory is 46% of systems versus 34% a year ago, about +119% in dollars. Foundry dollars still grew (about +24%) and are the same size as memory. The mix fell because memory grew faster, not because foundry rolled over. China of $2,886 million grew +26.4% and is 35% of sales, down from 43% a year ago. Ex-China revenue grew about +75%. Korea of $1,634 million grew +110.5%. The acceleration is memory geography, not a China restock.
New guidance is a constrained step, not a demand ceiling. December midpoint is revenue $8.6 billion and non-GAAP EPS $2.33, gross margin 52.5%, operating margin 40.5%. Versus last December that is about +60.9% revenue and +83.5% EPS. FMP sits a touch above both midpoints and inside the range. Management said the revenue guide would have been a bit higher without supply constraints, and would not size the haircut. Calendar 2027 wafer-fab equipment is a first figure of more than $200 billion, capped by clean rooms and components, with DRAM still ahead of foundry logic. There is no fiscal 2027 P&L guide.
Tone, breaks, and what is next. The call was specific on the etch and deposition mechanism (3D DRAM, NAND tiering, packaging) and refused the 2027 quantities. Three statement breaks: domestic China called “exactly as expected” after a decline guide, NAND “about +40%” sequential that is +52% on the mix math, and June operating expense mislabeled $1 billion. The January 27, 2027 call is where the WFE number gets a sharper pencil. KLA (October 28) and Applied Materials (November 12) are the near-term test of whether the constraint got worse.
SharePoint, OneNote, Outlook, and Excel were not available. Consensus is FMP, pulled October 26, 2026, not Visible Alpha or Bloomberg. The FMP file still held the pre-print September average, so the December street bar is not a confirmed post-call revision tape.
| Revenue | $8,219M · +54.4% YoY · +0.95% vs FMP | Non-GAAP diluted EPS | $2.20 · +74.6% YoY · +0.92% vs FMP |
| GAAP diluted EPS | $2.19 · +76.6% YoY · $0.01 under non-GAAP | Systems / CSBG | $5,733M +61.6% · $2,487M +40.0% |
| Non-GAAP gross margin | 52.4% · +180 bp YoY · +40 bp vs guide | Non-GAAP operating margin | 40.2% · +520 bp YoY · first print above 40% |
| EBITDA | $3,406M · margin 41.4% · +517 bp YoY | Memory / foundry mix | 46% / 46% of systems · memory $ about +119% YoY |
| China / Korea | $2,886M +26.4% · $1,634M +110.5% | Non-GAAP tax rate | 16.5% · +230 bp YoY · was 11.0% in June |
| December guide vs FMP | $8.6B / $2.33 vs $8.644B / $2.349 | Beat record | Revenue 12/12 · non-GAAP EPS 12/12 |
Systems are about 70% of sales and the earnings engine. Customer support is the installed-base line. Memory is where the incremental dollar is. China is still more than a third of sales, and it diluted the growth rate.
| Metric | FQ2'25 | FQ3'25 | FQ4'25 | FQ1'26 | FQ2'26 | FQ3'26 | FQ4'26 | FQ1'27 |
|---|---|---|---|---|---|---|---|---|
| Revenue ($M) | 4,376 | 4,720 | 5,171 | 5,324 | 5,345 | 5,841 | 6,722 | 8,219 |
| Revenue YoY | +16.4% | +24.4% | +33.6% | +27.7% | +22.1% | +23.8% | +30.0% | +54.4% |
| Rev accel (bps) | −326 | +799 | +915 | −584 | −560 | +162 | +623 | +2,439 |
| Systems ($M) | 2,626 | 3,035 | 3,438 | 3,548 | 3,357 | 3,731 | 4,250 | 5,733 |
| Systems YoY | +14.2% | +26.7% | +58.4% | +48.3% | +27.9% | +22.9% | +23.6% | +61.6% |
| CSBG ($M) | 1,750 | 1,685 | 1,734 | 1,777 | 1,987 | 2,111 | 2,472 | 2,487 |
| CSBG YoY | +20.0% | +20.5% | +1.9% | +0.1% | +13.5% | +25.3% | +42.6% | +40.0% |
| Non-GAAP GM | 47.5% | 49.0% | 50.3% | 50.6% | 49.7% | 49.9% | 52.0% | 52.4% |
| NG GM YoY (bp) | −10 | +30 | +180 | +240 | +220 | +90 | +170 | +180 |
| Non-GAAP OM | 30.7% | 32.8% | 34.4% | 35.0% | 34.3% | 35.0% | 38.4% | 40.2% |
| NG OM YoY (bp) | +70 | +250 | +370 | +410 | +360 | +220 | +400 | +520 |
| EBITDA ($M) | 1,430 | 1,659 | 1,840 | 1,931 | 1,914 | 2,163 | 2,633 | 3,406 |
| EBITDA margin | 32.7% | 35.1% | 35.6% | 36.3% | 35.8% | 37.0% | 39.2% | 41.4% |
| NG diluted EPS | $0.91 | $1.04 | $1.33 | $1.26 | $1.27 | $1.47 | $1.82 | $2.20 |
| NG EPS YoY | +21.0% | +33.5% | +63.4% | +46.5% | +39.6% | +41.3% | +36.8% | +74.6% |
| EPS accel (bps) | −454 | +1,249 | +2,989 | −1,688 | −695 | +179 | −450 | +3,776 |
The FQ1 stack is the comparison that matters. Revenue went +19.7% → +27.7% → +54.4%. Operating-margin expansion widened +80 bp → +410 bp → +520 bp. EPS went +25.5% → +46.5% → +74.6% and is now running about 2,020 bp ahead of revenue. Gross-margin expansion is still positive and is no longer the line that is accelerating. The leverage is below gross margin.
| Systems mix and geography | FQ2'25 | FQ3'25 | FQ4'25 | FQ1'26 | FQ2'26 | FQ3'26 | FQ4'26 | FQ1'27 |
|---|---|---|---|---|---|---|---|---|
| Memory % of systems | 50% | 43% | 41% | 34% | 34% | 39% | 46% | 46% |
| Memory $ YoY (calc.) | +19.0% | +23.8% | +80.4% | +44.0% | −13.0% | +11.5% | +38.7% | +118.6% |
| NAND % of systems | 24% | 20% | 27% | 18% | 11% | 12% | 23% | 26% |
| NAND $ YoY (calc.) | +61.2% | +20.7% | +151.6% | +142.6% | −41.4% | −26.3% | +5.3% | +133.4% |
| DRAM % of systems | 26% | 23% | 14% | 16% | 23% | 27% | 23% | 20% |
| DRAM $ YoY (calc.) | −4.2% | +26.7% | +16.7% | −1.2% | +13.1% | +44.3% | +103.1% | +102.0% |
| Foundry % of systems | 35% | 48% | 52% | 60% | 59% | 54% | 44% | 46% |
| Foundry $ YoY (calc.) | +5.2% | +38.2% | +91.6% | +117.0% | +115.6% | +38.3% | +4.6% | +23.9% |
| China ($M) | 1,343 | 1,467 | 1,837 | 2,283 | 1,859 | 1,997 | 1,722 | 2,886 |
| China YoY | −10.1% | −8.7% | +21.9% | +46.5% | +38.4% | +36.1% | −6.3% | +26.4% |
End-market dollars are the cited percent times cited systems revenue. They are not a separate fundamental. NAND plus DRAM equals memory. September memory dollars of about $2,637 million match the CFO’s “up about 120%.” Foundry dollars are about the same $2,637 million, on a mix that fell from 60% to 46%. Logic and other is 8% of systems and about +116% in dollars. It is real and it is still small.
Five-year shape. The annual slope is still only a modest steepening. The quarterly run-rate has left it behind. FQ1'27 revenue is already 35% of all of fiscal 2026 revenue.
| Fiscal year | FY2022 | FY2023 | FY2024 | FY2025 | FY2026 |
|---|---|---|---|---|---|
| Revenue ($M) | 17,227 | 17,429 | 14,905 | 18,436 | 23,233 |
| Revenue YoY | +17.8% | +1.2% | −14.5% | +23.7% | +26.0% |
| Systems ($M) | 11,322 | 10,696 | 8,922 | 11,491 | 14,885 |
| CSBG ($M) | 5,905 | 6,733 | 5,984 | 6,944 | 8,347 |
| China ($M) | 5,412 | 4,463 | 6,294 | 6,205 | 7,860 |
| Non-GAAP GM | 45.7% calc. | 45.3% calc. | 48.2% | 48.8% | 50.6% |
| EBITDA ($M) | 5,716 | 5,517 | 4,624 | 6,287 | 8,641 |
| NG EPS, split-adj. | $3.31 calc. | $3.41 calc. | $3.03 calc. | $4.14 | $5.82 |
| NG EPS YoY | +21.5% | +3.0% | −11.1% | +36.6% | +40.6% |
| Free cash flow ($M) | — | — | 4,256 | 5,414 | not returned |
Pre-split non-GAAP EPS, divided by 10 for the split-adjusted cells: FY2022 $33.11, FY2023 $34.10, FY2024 $30.30. FY2022 and FY2023 non-GAAP gross margin is non-GAAP gross profit divided by revenue: $7,868 million and $7,887 million. FY2025 free cash flow was +27.2% versus FY2024. That is the only clean annual cash comparison in the pull. Do not read the P&L acceleration as a cash-conversion conclusion. One September free-cash-flow figure was returned, $3,240 million. The quarterly series is missing other quarters in this window and is not trended here.
FMP fiscal 2027 averages, not a Lam guide and not Visible Alpha: revenue $35.41 billion, EPS $9.69, EBITDA $14.37 billion. Against fiscal 2026 that is about +52% revenue, +67% EPS, and +66% EBITDA. It extends the annual acceleration. It was not refreshed from Bloomberg.
What the table shows. Revenue growth slowed, based at +22.1% in FQ2'26, and then broke out. A 54% grower that is still speeding up is a different setup from a 30% grower that is fading. The quality check is that foundry dollars did not shrink and GAAP EPS grew with non-GAAP EPS. The open question is duration of the memory dollar, not whether September was a China quarter.
Consistent beater. The EPS surprise just compressed to the smallest print in the window. Revenue beats have settled near +1%. The business trajectory accelerated. The surprise did not. A beat is any positive surprise versus the FMP earnings-history estimate. No quarter in the last 12 is a dead zero or a miss.
| Versus FMP | FQ2'25 | FQ3'25 | FQ4'25 | FQ1'26 | FQ2'26 | FQ3'26 | FQ4'26 | FQ1'27 |
|---|---|---|---|---|---|---|---|---|
| Revenue vs FMP | +1.4% | +1.7% | +3.3% | +1.9% | +2.1% | +1.6% | +0.9% | +1.0% |
| NG EPS vs FMP | +3.6% | +4.0% | +9.9% | +3.3% | +8.5% | +8.1% | +7.7% | +0.9% |
| Revenue YoY | +16.4% | +24.4% | +33.6% | +27.7% | +22.1% | +23.8% | +30.0% | +54.4% |
| NG EPS YoY | +21.0% | +33.5% | +63.4% | +46.5% | +39.6% | +41.3% | +36.8% | +74.6% |
Heatmap actuals, non-GAAP EPS on the series matched to FMP: $0.91, $1.04, $1.33, $1.26, $1.27, $1.47, $1.82, $2.20. This quarter’s FMP bar was $8,141.8 million and $2.18. The finer analyst-estimates EPS mean of $2.182 still leaves a beat of about +0.8%.
The four quarters before that window, so the 12-quarter record is auditable. Pre-split EPS is divided by 10. FMP’s FY2024Q3 EPS estimate is stored pre-split at $7.30. Using that raw figure against a split-adjusted actual would fake a miss. Split-adjusted it is $0.73.
Rates. Last 12 quarters (FQ2'24 through FQ1'27): revenue 12/12, non-GAAP EPS 12/12. Last four quarters: 4/4 on both. The sole revenue miss in the prior window, FQ1'24 at −0.9%, has rolled off.
Magnitude is deteriorating on the scored line, not on the business. Average revenue surprise fell from +2.08% in the middle four quarters to +1.38% in the last four, and the last two prints are +0.86% and +0.95%. Average EPS surprise over the last four is still +6.31% only because it includes three prints near +8% and then +0.92%. The latest dollar beat is $0.02, against $0.10, $0.11, and $0.13 on the prior three. The percent was going to narrow as the denominator rose. A drop from +8% to +1% is not that. It is the tax rate moving from 11.0% to 16.5%, which the CFO had already flagged, plus a gross-margin guide that started at the margin just printed and was cleared by 40 bp.
Versus Lam’s own September guide, the quarter was also a narrower clearance than June. Revenue landed $8,219 million versus a $8.1 billion midpoint, +1.5%, inside the high end. Non-GAAP EPS landed $2.20 versus $2.15, +$0.05, inside a $2.30 high. June’s gross margin, operating margin, and EPS had cleared that quarter’s high end. September did not. The company is still hitting above the midpoint. The amount of upside versus the guided range narrowed. FMP had already sat 0.5% above the revenue midpoint and 1.4% above the EPS midpoint before the print, so most of a normal clearance was in the street number.
The operating story under the small beat is still improving year over year. Non-GAAP operating income grew +77.3% to $3,301 million from $1,862 million. EPS grew less than operating income because of tax, not because of a buyback. Diluted shares were 1,255.8 million versus 1,269.3 million, −1.1%.
Lam guided only the next quarter, fiscal Q2 FY2027 (December 2026). Daloopa stores the outlook on the quarter in which it was issued, so the December outlook sits on fiscal 2027Q1. There is no second pass, and there is no fiscal 2027 revenue, EPS, margin, or EBITDA guide. Guidance on the call is non-GAAP.
Ranges are ± $400 million, ± 1 point of gross margin, ± 1 point of operating margin, and ± $0.15 of EPS. The YoY compare is fiscal Q2 FY2026, not the quarter just printed: revenue $5,345 million, gross margin 49.7%, operating margin 34.3%, EPS $1.27. At the ends of the range, revenue growth is about +53% to +68% and EPS growth is about +72% to +95%. Only the midpoint and the high end extend the revenue acceleration. The low end is about +53.4%, a small deceleration versus +54.4%.
Lam had not guided December before. The waterfall below is the step from the last numerical outlook (the September guide) to the new midpoint, then to FMP. It is not a same-period revision.
Other forward statements, with the qualifier from the call where Daloopa stores only the point.
- Calendar 2026 WFE is now the mid $150 billions. Daloopa stores $150 billion, the same point as July’s $150 billion. The change is the word “mid,” not the stored point. The path was $135 billion, then $140 billion, then low $150 billions, then mid $150 billions. Each step was labeled.
- Calendar 2027 WFE is a first figure, more than $200 billion. Growth ranked DRAM, then leading-edge foundry logic, then NAND. The rank has not changed in 90 days. Unconstrained demand would be higher. No dollar gap. A sharper figure is planned for the January call.
- Calendar 2026 systems growth, including the December midpoint, is well over the stored 40%. CSBG calendar 2026 is more than 30%. Calendar 2027 CSBG is over 15%. Support growth is being guided to cool. Systems carry the “well over 40%.”
- Advanced packaging growth of 70% for calendar 2026 was reaffirmed. Next year, “nicely,” with no number.
- Gross margin over several years is aimed at the mid-50s. Near term it should bounce around the current level. December is guided up 50 bp despite a customer-mix headwind. The tax rate is expected to stay near 16.5%.
- China should be a significantly lower percentage of December revenue than September’s 35%, driven by domestic customers. Multinationals in China are expected to stay reasonably strong. Nothing in December was called episodic.
- Ten new tier-1 fabs layer in across calendar 2027. Demand visibility was described through 2028. The limit is clean-room timing and components, not a shorter list of projects.
Replacing only the December street average with the $8.6 billion midpoint, and leaving March ($9.014 billion) and June ($9.523 billion) FMP averages untouched, puts fiscal 2027 near $35.36 billion. That is about 0.1% under the $35.41 billion annual average. The December guide does not reset the year. The supply spill management described would land in March, and it was not sized.
Guidance behavior. The team still hits what it frames, with one nuance. WFE has been walked up all year, and quarterly actuals have come in above the midpoint. The September beat did not clear the top of the range the way June’s margin and EPS did. Confidence on demand is higher. Confidence that every dollar of that demand ships in the quarter being guided is more qualified than it was in July. Tim Archer separated the current supplier tightness from the COVID shortages and said Lam should be caught up shortly. Doug Bettinger refused a dollar figure after an analyst suggested $300 million to $400 million.
Revenue year-over-year has re-accelerated for three straight quarters, from a FQ2'26 local trough of +22.1% to +54.4%. Non-GAAP EPS did not turn with it. Its growth rate kept fading into FQ4'26 and only inflected this quarter.
The window is a W, and every print is still positive. It opens at +16.4% in FQ2'25, still decelerating. Two accelerations carry it to a first-leg peak of +33.6% in FQ4'25. Two decelerations give most of that back, to +22.1% in FQ2'26. That quarter is the turn. It is not the eight-quarter low. Three accelerations follow: +162 bp, +623 bp, then +2,439 bp.
EPS spiked to +63.4% in FQ4'25, then decelerated for most of fiscal 2026. A +179 bp bounce in FQ3'26 did not hold. The EPS inflection is this quarter only. The FQ4'26 EPS slowdown was a base effect, not margin compression. In that same quarter non-GAAP operating margin expanded +400 bp to 38.4% from 34.4%, and non-GAAP operating income grew +45.1% to $2,581 million from $1,779 million. FQ4'25 EPS had just grown +63.4%, so FQ4'26 was dividing by a spiked base.
The December guide midpoint, not a ninth reported column, is +60.9% revenue and +83.5% EPS versus fiscal Q2 FY2026, or +653 bp and +886 bp of further acceleration if the midpoint is hit. Management said the revenue guide would have been a bit higher without the supply constraint. FMP’s $8.644 billion and $2.349 sit a hair above those midpoints.
Inflection points to keep.
- FQ2'26 revenue-rate trough at +22.1%, the turn, not the low of the window.
- FQ4'25 first-leg peak at +33.6% revenue and +63.4% EPS, the base that made the next EPS year look slower.
- FQ4'26 EPS-rate trough at +36.8%, while operating margin was still expanding.
- FQ1'27 joint breakout: revenue +54.4% and EPS +74.6%, operating margin through 40%.
- December midpoint, if hit, extends both rates. The low end of revenue does not.
What would falsify the inflection is a December print at the low end of the range that does not show the slipped shipments in March, or a China mix step-down that takes gross margin through the bottom of the 51.5% guide. Neither is the base case in the language on this call.
The street is underwriting the constrained December guide. It is not underwriting unconstrained 2027 WFE, the Aether raise, or an SLC mix shift. FMP December revenue of $8.644 billion and EPS of $2.349 sit inside $8.6 billion ± $400 million and $2.33 ± $0.15. Public write-ups on October 22–23 cited a bar nearer $8.5 billion and about $2.30 to $2.31. Either vintage is inside the band. Consensus is not paying, inside this quarter’s number, for the demand story management spent the hour on.
| Catalyst | When | What management said | What the December number does |
|---|---|---|---|
| Supply haircut | Now through Jan 27 | Guide would have been a bit higher. Slip lands in March. Not a COVID shortage. No dollar. | Street is on the constrained midpoint, not a make-up. |
| 2027 WFE | Sharper pencil Jan 27 | More than $200B, supply-capped. DRAM, then foundry logic, then NAND. Ten tier-1 fabs across the year. | Not in the $8.6B. SEMI’s published path reaches $200B in 2028, not 2027. |
| China mix | December quarter | Meaningfully lower percent, domestic-led. Multinationals in China stay reasonably strong. | Already required for $8.6B to work. A BIS snap-back is a different cut and is not in FMP. |
| 3D DRAM vs 4F² | Multi-year | A skip from 6F² to 3D raises etch and deposition intensity faster. Extending today’s node also helps Akara and Aether. | The clearest management-versus-street gap. Not a December plug. |
| SLC NAND | 2027 and after | Each 1 point of bits moving to SLC-class needs about 2% more wafer capacity. Mix refused. | No SLC adder. The +133% NAND systems print is already in the base. |
| Aether | Through the decade | About $2.5B cumulative, roughly $1B above the prior expectation. In HBM high-volume manufacturing. Call figure, not a Daloopa actual. | Not in the $8.644B. December gross margin is the 52.5% guide, not the mid-50s. |
| CSBG | CY26 and CY27 | September +40%. Guide cools to more than 30% this year and over 15% next year. Ex-China mature spend is a 2027 item. | Do not keep a 40% support rate in the model. |
Where management is ahead of the street. Three statements are being repeated, and none of them is in the December consensus. First, the $200 billion-plus 2027 WFE figure is a supply number, and unconstrained demand is higher. Second, 3D scaling is good for Lam even when a customer skips 4F². The questioner said the shares were read the other way. Third, Aether, panel-level share, and SLC wafer math are real product catalysts sitting outside $8.6 billion. Consensus enthusiasm has already been spent moving the near-term model up to a guide that only barely beat September. The residual argument is the one management is making.
What would change the December number before January 27. KLA on October 28 and Applied Materials on November 12, if either says the constraint got worse. A BIS Affiliates Rule snap-back on November 10, which would be a service cut, not the domestic timing call already in the guide. Memory-customer capex commentary, which feeds the 2027 DRAM rank and should not be used to lift December Lam revenue above the guide unless it comes with a shipment comment.
Peers named here are from Daloopa industry 312 (KLAC company id 111, AMAT company id 12), plus ASML (id 890) on the lithography side of the same constraint. No shipment alternative data. Snowflake and Carbon Arc were not connected.
Well answered on the mechanism. Deflected on the 2027 quantities. Tim Archer’s line, repeated until the queue closed, is that deposition and etch intensity rises whether DRAM goes 3D, NAND splits into SLC, TLC, and QLC, or foundry adds backside power and packaging. The street kept asking for the number that turns that mechanism into a model. Thirteen analysts got on. The tally is 15 well answered and 10 deflected, out of 25 scored items. The deflections cluster on unconstrained WFE, the China percent, the supply-chain dollar haircut, SLC mix, packaging content per wafer, CSBG margins, and the accelerated-share-repurchase size.
| # | Analyst | Ask | Result |
|---|---|---|---|
| 1a | Arcuri, UBS | Dollar size of the December supply haircut | Deflected |
| 1b | Arcuri, UBS | 3D DRAM versus 4F² | Answered |
| 2a | Weathers, Deutsche Bank | What relaxes the constraint; co-investment | Answered |
| 2b | Weathers, Deutsche Bank | NAND upgrade inning, 384-layer, SLC | Answered |
| 3a | Sankar, TD Cowen | Is December WFE episodic, or inflated by price | Answered |
| 3b | Sankar, TD Cowen | Quantify the December China drop | Deflected |
| 4a | Sur, JPMorgan | CSBG drivers and CSBG margin | Deflected |
| 4b | Sur, JPMorgan | Mature demand outside China, this year | Answered |
| 5a | Pajjuri, RBC | What drives 2027 outgrowth versus WFE | Answered |
| 5b | Pajjuri, RBC | Gross margin once China mix normalizes | Answered |
| 6a | Arya, Bank of America | What unlocks WFE above the ~30% 2027 start | Deflected |
| 6b | Arya, Bank of America | ASR size, start, and end | Deflected |
| 7 | Malik, Citi | Panel ramp, and whether DRAM’s lead has narrowed | Answered |
| 8 | Schneider, Goldman Sachs | What moves 2027 WFE up or down | Answered |
| 9a | Rasgon, Bernstein | Unconstrained 2027 WFE versus the $200B floor | Deflected |
| 9b | Rasgon, Bernstein | NAND upgrades versus new tools; flat CSBG | Deflected |
| 10 | Curtis, Jefferies | Is SLC real demand; packaging growth next year | Answered |
| 11a | Brett, Morgan Stanley | Foundry SAM per wafer versus Analyst Day | Deflected |
| 11b | Brett, Morgan Stanley | Does hybrid bonding help packaging share | Answered |
| 12 | Rakesh, Mizuho | SLC bit mix, and content per 100,000 CoWoS or EMIB-T wafers | Deflected |
| 13 | Chin, Stifel | Inventory turns, and whether the Aether raise is new customers | Answered |
The numbers the hour sat on, and then left: revenue $8,219 million and non-GAAP EPS $2.20 were not the debate. Inventory turns rose to 3.4x from 2.6x. Deferred revenue was $2,660 million, −4.0% versus $2,770 million. Bettinger said turns should stay in a normal band and that the Aether raise is both new customer adoptions and volume where it is already process of record. He would not size the repurchase that is still executing in December. About $1 billion of repurchases and $325 million of dividends in September, and about $3 billion left on the authorization, are call figures, not a Daloopa series in this pull.
What to do with the deflections. Do not fill the blanks. The model that is honest to the call uses the mix math for NAND, holds China as “meaningfully lower” without a fake percent, leaves the supply haircut inside the ±$400 million, and treats more than $200 billion as a constrained floor. The mechanism answers are usable. The quantities are not.
Three breaks. Two are high. One is a period label. Lam is otherwise consistent on the points it will put in the script: record September revenue and margins, memory at 46% of systems, CSBG near $2.5 billion and about +40% year over year, packaging still +70% for calendar 2026, and a preliminary 2027 WFE above $200 billion that they label as constrained and early.
July’s latter-half 2026 composition was specific: indigenous and domestic customers down, multinationals in China up, with lumpiness reserved for the total China percentage. October’s prepared remarks said China was 35% of revenue, up from June’s 26%, with strong investment from both domestic and multinational customers. In the Sankar exchange he said it was not a surprise, it was exactly as expected, and that domestic customers in September were strong. December, he said, will be meaningfully down, driven by those same domestic customers.
Both descriptions cannot be the July expectation. Lumpiness does not convert a forecast of domestic decline into a forecast of domestic strength. The December guide quietly returns to the July composition, which is how you can tell September domestic strength was the break.
Filings, year over year: September China $2,886 million, +26.4% versus $2,283 million. Mix 35% versus 43% a year ago. June was $1,722 million, −6.3% year over year, at 26% of sales. The year-over-year path is up, down, up. That is lumpiness. It is not evidence that domestic decline was the expected September outcome. Do not model a flattish China year off the July sentence.
The script said non-volatile memory was 26% of systems, up from 23% in June, and that NAND revenue increased approximately 40% sequentially. Systems revenue was up about 35% sequentially. Those cannot be the same NAND.
June systems $4,250 million at 23% is about $977 million. September systems $5,733 million at 26% is about $1,491 million. That is +52.5% sequential and about +133% year over year versus an 18% mix on $3,548 million. A 40% step off the June base would have been about 23.9% of September systems, not 26%. Memory dollars overall do reconcile with the script’s “up 120%.” The break is isolated to the NAND 40%.
Rasgon noted customer support was roughly flat and asked whether upgrades drove the NAND number. Bettinger said new equipment and upgrades were both meaningful. He did not reconcile 40% to 26%. Customer support of $2,487 million was only about +0.6% versus June’s $2,472 million. A flat services line cannot pull a +52% systems result down to +40%. Use the mix. The 40% understates how much of September was NAND systems.
July: operating expenses in the June quarter were $916 million, up from $866 million. October: operating expenses for the June quarter were $1 billion, up from $916 million. The second clause gives the resolution. The $916 million base is June, so the $1 billion figure is September with the wrong name on it.
Non-GAAP operating expense: June $916 million, September $1,004 million, +20.7% versus $832 million a year ago. June GAAP operating expense was $965 million, not $1 billion. September expense grew much more slowly than September revenue. Keep June at $916 million. Do not restate it.
Checked and not counted: the WFE ladder (each raise was labeled), the CSBG framework (the old high-single-digit model was retired in public), the gross-margin “level out” line (replaced, with the drivers named), China as a percent of Lam coming down for the year (the mix-down call is not broken; the domestic-versus-multinational composition is, and that is card 1), 4F² versus 3D DRAM (a timing shift, and he did not say 3D is bad for Lam), and the $200 billion figure versus “too early to be precise” (reconciled on the call as a constrained starting point).
No view on policy rates, CPI, or consumer confidence. The only consumer reference is demand: AI adoption by enterprises and consumers sits under the WFE outlook. Higher interest income was one reason other income flipped to about $16 million of income. That is a P&L description, not a rate view. The 16.5% tax rate is tax and geographic mix, not inflation.
The industrial tone is early in a multi-year build, and the constraint is capacity, not orders. Calendar 2026 WFE is the fourth raise of the year, to the mid $150 billions, because customers squeezed clean rooms and pulled projects forward. Calendar 2027 is more than $200 billion and still constrained. Utilization is high enough that spares are “humming,” and taking tools down for NAND upgrades is itself a scheduling problem. Leading-edge tightness is spilling into analog, power, and microcontrollers, but mature spending outside China is “probably not” meaningful in 2026. Bettinger expects it to be strong in 2027.
| Who | How they entered | Read-through | Confidence |
|---|---|---|---|
| Micron and other memory makers | Analyst, not Lam. Same-day 3D-over-4F² comment. | Better for etch and deposition if the industry skips 4F². Either path raises Lam intensity. Korea $1,634 million, +110.5%, is the footprint. Lam did not split Samsung from SK hynix. | Direct on the mechanism. Inferred on the name. |
| AMAT, TEL, KLAC, ASML | Not named. Asked as “front-end equipment companies.” | Do not read the WFE raises as peer price inflation. The clean-room and component limit is shared. A December slip at Lam is timing, not a Lam-only bottleneck. Aether’s lower exposure dose is better for scanner productivity, not a displacement of the scanner. | Direct on the industry constraint. Moderate on any one peer. |
| Component suppliers | Unnamed. Packaging subsystems called out. | Better for well-capitalized etch, deposition, and packaging suppliers. Worse only for a supplier that cannot add capacity. That is what pushes Lam shipments from December into March. No co-investment amount. | Direct on the condition. No ticker. |
| China domestic vs multinationals in China | Prepared remarks, then Sankar. No customer named. | Near-term worse for domestic China wafer-fab spending as a December mix item. Not described as a cancel. Relatively better for multinational fabs inside China. China is already the slow geography at +26.4% against company revenue +54.4%. | Direct on the mix. Not a named share shift. |
| Foundry logic, including Terafab | Analyst named Terafab. Management did not. | Does not upgrade 2027 foundry logic versus DRAM. Public fab headlines were already inside the $200 billion starting point. Foundry mix of systems is down 1,400 bp and foundry dollars are still up. One leading customer took tool-of-record wins on four backside-power products. Do not assign that win to TSMC or Intel. | Direct on the rank. The customer is unnamed. |
| Packaging: CoWoS, EMIB-T, panel | Process names from the analysts. More than 20 panel customers, unnamed. | Better for Lam as packaging grows, whether the flow is CoWoS, EMIB-T, or hybrid bonding. Lam does not do the bond. No wafer-based dollar. Not a 2027 WFE swing by itself. Management still calls panel early. | Direct that the roadmap is active. Unquantified. |
| Mature analog, power, MCU | Sur. No ticker named. | A 2027 read, not a 2026 one, once China is stripped out. Reliant was strong in September on mature-node spending. Not evidence those factories are already spending heavily this year ex-China. | Direct on the timing. No name. |
Geography, same September quarters. Taiwan $1,482 million was +44.3% versus $1,027 million, mix 18% versus 19%. The United States was $663 million versus $305 million, +117.6%, mix 8% versus 6%. Korea mix was 20% versus 15%.
What not to over-read. No company was singled out as losing share to Lam. “Third consecutive year of WFE outperformance” is a Lam statement about Lam, not a named take from Applied Materials or Tokyo Electron. SLC, high-bandwidth flash, CoWoS, and EMIB-T were left unquantified on purpose. The only numeric NAND translation management would give is mechanical: about 2% more wafer capacity per 1 point of bits moving to SLC-class devices.