Lam’s beat was not the story; the China haircut is smaller than the foundry mix shift is worth

Lam Research did enough in the September quarter to keep the stock argument centered on sustainability rather than catch-up. What was priced in was a company already benefiting from WFE recovery, with revenue expected at $5,223.1 million and EPS expected at $1.22. What actually surprised was narrower: revenue came in at $5,324.2 million, a +1.9% surprise, and EPS came in at $1.26, a +3.3% surprise. That is not the kind of beat that should reset a thesis by itself. The variant perception is that the market may treat the newly quantified China restriction as an incremental reason to fade the cycle, when the print shows Lam is entering the restriction from a materially stronger mix position: foundry reached 60% of systems revenue, while memory fell to 34% because of timing rather than a demand break. The quarter therefore argues for a higher quality revenue base, not simply a higher revenue number.

The distinction matters because this was not a low-quality upside quarter built on a transient shipment pull-forward. Revenue has moved from a trough of $3,207.3 million to $5,324.2 million across the displayed cycle, while gross margin expanded from 45.5% to 50.4%. The shape of that recovery is the key, because Lam is no longer just regaining lost volume; it is holding margin above the late-cycle level while absorbing mix changes and export-rule friction. The Street beat was small enough to be dismissed, but the financial trajectory is not. The company has now crossed the $5 billion quarterly revenue level on the reported basis used in the call, and management framed that threshold as more than a one-quarter peak. Tim Archer’s wording matters because it commits to duration, not just the September print: “Inclusive of our guidance for the December quarter, we expect to close calendar 2025 with three consecutive quarters of greater than $5 billion in revenue.”

That durability is also why the China issue should be sized rather than treated as an open-ended overhang. Archer said the December guide includes “roughly a $200 million revenue impact from the recently announced 50% affiliate rule restricting shipments to certain domestic China customers.” He then put a calendar-year number on the constraint: “Currently, we expect this rule to impact our calendar year 2026 revenues by approximately $600 million.” Those two figures are the crux of the debate. A $200 million December hit is meaningful against company guidance of $5.2 billion, plus or minus $300 million, but it is not a thesis-breaking shock if the company can still guide around the $5 billion level. The $600 million calendar-year 2026 figure is the number PMs should underwrite against foundry and memory mix, because it converts export control from a qualitative fear into a quantifiable revenue headwind.

The guide itself is conservative enough to leave room for the market to miss the mix improvement. Doug Bettinger guided December revenue to $5.2 billion, plus or minus $300 million, and EPS to $1.15, plus or minus $0.10, based on approximately 1.26 billion shares. On its face, that EPS guide sits below the Street-comparison EPS actual of $1.26, so the initial read can be that Lam beat and guided down. But the better interpretation is that management is embedding both the China restriction and the timing-driven memory pause while still sustaining a revenue level that management says keeps calendar 2025 at “three consecutive quarters of revenue above $5 billion.” The company’s reported-basis call numbers should not be conflated with the Street-comparison figures in the print: the print uses EPS actual $1.26 for the beat, while Archer described company accounts with record revenues of $5.3 billion and gross margin of 50.6%. The apparent basis differences do not alter the conclusion: the surprise was not magnitude, it was resilience after explicit China haircut.

The capacity mix explains why the revenue floor matters more than the December EPS optics. Foundry rose to 60% of systems revenue from 52%, while memory declined to 34% from 41%. That is a large rotation inside systems revenue, and it has second-order implications for customer read-through. For TSMC, Intel, and Samsung, the 60% foundry share points to continued demand for conductor and dielectric etch, CVD/ALD deposition, and wet clean, all areas where Lam is exposed. For memory customers, the message is more nuanced: nonvolatile memory fell to 18% of system revenue from 27%, while DRAM increased to 16% from 14%. That split is more favorable for Micron and SK Hynix DRAM spending than for Kioxia and 3D NAND timing, but it is not a memory collapse because DRAM share increased even as total memory share declined.

The China geography and customer implications also cut against a simplistic bearish read. Domestic China restrictions matter most for SMIC, where Lam’s etch and deposition exposure is explicitly export-control constrained, and management’s quantified $200 million December impact gives investors a level to monitor. But the geographic mix already showed the pull of non-China demand: Taiwan was 19% and flat sequentially, while Korea was 15% and down from 22% because of customer investment timing. That mix is consistent with the segment data: foundry strength, Korea-linked memory timing, and China-specific restriction. Suppliers should read the quarter as a better demand signal for tool-complexity content than for broad unit acceleration. Ichor Systems gas delivery, Ultra Clean Holdings chambers and subsystems, MKS Instruments equipment, Advanced Energy equipment, Horiba sensors and MFCs, and Ferrotec quartz and ceramic components all lever into Lam’s etch and deposition shipments, but the December $200 million China headwind argues against extrapolating September demand in a straight line for export-sensitive configurations.

The margin story is where the bull case has the most discipline, because it is not dependent on claiming every revenue dollar is equally profitable. Gross margin on the quarterly history stands at 50.4%, up from 48.0% a year earlier, despite the systems mix rotation and despite service revenue being a meaningful piece of the model. Bettinger said the customer support business generated approximately $1.8 billion in September-quarter revenue, slightly higher sequentially and year over year. That services base matters because it provides a stabilizer when system timing shifts between foundry, DRAM, and nonvolatile memory. Operating expense was $832 million, up from $822 million, so the margin thesis is not being manufactured by a dramatic cost cut. It is being carried by scale, mix, and a services layer that is large enough to matter but not large enough to mask systems-cycle direction.

Capital return adds support, but it should not be the central reason to own the stock after this print. Lam ended September with $6.7 billion of cash and cash equivalents, up from $6.4 billion at June quarter-end, and had $6.5 billion remaining on its Board-authorized repurchase plan. The company has also repurchased nearly 30 million shares year to date at an average price of a little more than $88 per share. Those figures matter because they set a floor under per-share math if revenue stays near management’s $5 billion-plus framework, but they do not offset a true WFE downturn. The print’s value is that it makes a downturn harder to argue from the available evidence: the top line beat, the gross margin stayed around the 50% level, and the China hit was sized rather than left open.

The competitive context also supports the view that Lam’s margin quality is being underappreciated. In the latest peer table, Tokyo Electron’s listed gross margin is 46.8% with revenue YoY of +10.6%, while Lam’s September-quarter history shows gross margin of 50.4% and revenue YoY of +27.7%. The point is not to compare currencies or declare share gain from one table; the point is that Lam is showing higher margin and faster year-over-year revenue growth than the most directly comparable WFE peer in the provided set. That relative shape is important for PMs because the bear case often treats wafer fab equipment as a synchronized beta trade. This print argues Lam is not merely riding the same slope; its etch/deposition exposure is participating in the specific parts of spending tied to foundry complexity and AI infrastructure.

The AI infrastructure linkage is not just marketing language in this call, but it still needs to be handled with discipline. Archer estimated “roughly $8 billion of WFE spending for every $100 billion in incremental data center investment.” That statement is useful because it ties Lam’s demand case to a capital intensity conversion rather than to generic AI enthusiasm. It also helps explain why foundry reached 60% of systems revenue in the September quarter: the industry needs advanced logic capacity, and Lam’s tools are tied to the patterning, deposition, and clean steps that become more demanding at leading nodes. The risk is that the $8 billion conversion is an industry estimate, not Lam-specific revenue guidance. The actionable point is therefore not to add $8 billion to Lam’s model, but to recognize that the company’s foundry mix gives it a clearer route to data-center capex than a memory-only equipment supplier would have.

The call delivery was less promotional than the numbers might suggest, and that matters because tone often flags whether management is stretching the cycle narrative. The tone history shows sentiment fell to 0.23 from 0.29, while guidance_tone declined to 0.32 from 0.42. That softer delivery could look negative at first glance, but uncertainty also fell to 57.4 from 66.0, which is the more important signal after a quarter shaped by export controls. Management did not sound more euphoric; it sounded more bounded. The conflict is in qa_evasiveness, which rose by +25.1 call over call, so investors should not ignore the possibility that management had less room to answer around China and customer timing. Still, the simultaneous drop in uncertainty suggests the company had converted the biggest unknown into the $200 million and $600 million figures investors can now track.

That tone profile fits the print’s investment message: Lam is not asking investors to believe in an unbounded acceleration, but it is giving enough hard numbers to challenge the bear case. The most important hard numbers are the Street-comparison beat of +1.9% revenue and +3.3% EPS, the foundry share of 60%, and the calendar-year 2026 China impact of approximately $600 million. If the stock trades only on the December guide of $5.2 billion, plus or minus $300 million, the market may miss that the guide already embeds a roughly $200 million regulatory hit. If the stock trades only on China, the market may miss that memory weakness was described through timing and that DRAM share moved up to 16% even as nonvolatile memory fell. If the stock trades only on AI enthusiasm, the market may overpay for a broad theme; the better thesis is narrower and more defensible, that Lam’s mix and margin resilience justify looking through a quantified export-control drag.

What to watch next is therefore concrete. First, December-quarter revenue needs to land inside or above the $5.2 billion, plus or minus $300 million guide after absorbing the roughly $200 million China impact; a print below that range would break the thesis that the haircut is manageable. Second, foundry should remain the dominant systems segment after reaching 60%, because a sharp reversal without a corresponding DRAM or nonvolatile memory rebound would weaken the mix argument. Third, gross margin should stay near the 50.4% recent level, since the bull case depends on Lam holding economics while revenue remains above the $5 billion threshold management highlighted for calendar 2025. Finally, management’s next update on calendar-year 2026 China exposure must stay close to approximately $600 million; a material upward revision would turn a quantified overhang into a widening one.