Layer 2 · LRCX — Lam Research
One-line thesis
Lam Research owns the etch and deposition franchise that physically makes HBM through-silicon vias and 3D NAND stacks — the two highest-growth equipment lines in the cycle — and the Q3 FY26 record print plus the SK Hynix / Samsung HBM TSV capacity ramp through 2027 makes this the equipment-pair-trade alongside KLAC.
What Lam Research physically does
Lam makes the plasma-etch tools (the Kiyo, Flex, and Versys lines for conductor and dielectric etch) and the deposition tools (Striker ALD, Vector PECVD, ALTUS tungsten CVD) that perform two specific operations in semiconductor manufacture: removing precisely-defined volumes of material from the wafer surface (etch) and laying down precisely-controlled thin films of material on the wafer surface (deposition). At leading-edge nodes these steps run hundreds of times per wafer, and the binding-constraint volume for AI memory specifically comes from two product lines.
First, the deep-silicon etch tool — the Sense.i Plus Bosch-process etcher — is the principal piece of equipment that creates through-silicon vias (TSVs) for HBM. Every HBM die has thousands of TSVs running vertically through the silicon to connect adjacent layers in the stack; etching those vias to ~5µm diameter and ~50µm depth at 100,000+ vias per die per hour throughput is what Sense.i Plus does, and Lam holds approximately 80%+ market share in this specific equipment niche. The physics are not trivial. The Bosch process alternates between an SF6-based etch step and a C4F8-based passivation step at sub-second cycle times, producing scallop-free vertical sidewalls at aspect ratios exceeding 10:1. Doing this at production throughput on 300mm wafers carrying tens of thousands of vias each is the operational moat that has kept Lam dominant against TEL (which competes in the broader silicon-etch market but has not won meaningfully at the HBM-specific TSV step) and Applied Materials (which competes at logic etch but has minimal HBM TSV presence).
Second, the high-aspect-ratio etcher for 3D NAND vertical channels — the Flex-D family — performs a similar function at higher aspect ratios for 3D NAND vertical channels, where Lam holds 60-70% share against TEL. Modern 3D NAND structures at Kioxia, Western Digital BiCS, and Samsung V-NAND now stack 400+ layers vertically, with vertical channel aspect ratios exceeding 70:1 — physical requirements that have driven a multi-decade Lam-TEL technology race where Lam has held the share leadership since the 256-layer node.
The third critical line is atomic-layer deposition (ALD) for high-k metal-gate and capacitor films at advanced logic and DRAM nodes — the Striker family — where Lam competes with TEL and Applied Materials for share in a faster-growing TAM. ALD scales materially with both DRAM node migration (1z to 1a to 1b to 1c) and with leading-edge logic capacitor work, and Lam has been investing heavily in expanding the ALD portfolio.
The fourth segment is the broader conductor-etch (the Kiyo platform) and dielectric-etch (Flex platform) franchises, which serve every leading-edge logic node. As TSMC ramps N2 and Intel ramps 18A, these tools are pulled through capacity expansions at multi-billion-dollar order volumes per fab.
For an AI-stack investor: every HBM die ever made in production volume has passed through Lam Sense.i Plus etchers. The HBM cycle is mechanically pulling Lam revenue through every wafer increment, and the company's tool-utilisation data is one of the cleanest leading indicators of HBM capacity build-out in the entire equipment cohort.
The financial print
Q3 FY26 (reported April 22 2026, quarter ended March) printed record revenue of $5.84B (+24% YoY) and EPS of $1.47 against $1.36 consensus — beat by 8%. The company raised June-quarter guidance on AI memory etch demand. Gross margin held at 48.2% on a normalized basis. The Q3 result is meaningful because Lam's revenue base mix has flipped over 24 months — memory equipment now accounts for roughly 50%+ of the equipment line (versus historically 30-40%), with HBM-specific tool orders running well above pre-cycle peak.
The composition of the Q3 revenue is worth dissecting. System revenue grew 27% YoY to roughly $4.0B; customer support business operations (the installed-base service revenue) grew 18% YoY to approximately $1.85B. The service revenue — which historically runs at a 60%+ gross margin and is structurally non-cyclical — has been quietly compounding at mid-teens for several years and represents an increasing portion of the steady-state earnings power. Operating margin came in at 32.1% non-GAAP, the highest quarterly level since the 2022 cycle peak. The capital return announced alongside the print included continued aggressive buybacks (roughly $1.5B per quarter through 2026) plus a dividend hike that complements the operating leverage.
The stock at $285 is 1m +8.4%, 3m +24% (note: LRCX completed its 10-for-1 split in October 2024, so all share counts and historical comparison points are on the post-split basis). Q4 FY26 prints July 22. Consensus sits around $6.0-6.2B revenue and $1.60+ EPS. The same pattern as KLAC applies — the bar is high but the underlying demand visibility through 2027 is the strongest in two decades. The principal swing variable into the print is the China-revenue line: Lam's China revenue at Q3 ran approximately $1.5B (declining from the cycle peak of ~$2.2B in fiscal 2024) and the consensus FY27 model assumes that line stabilises rather than deteriorates further. Any further restriction tightening compresses that line on the next print.
Customer mix
SK Hynix is now one of the largest individual customers — likely 12-15% of revenue currently, driven by HBM3e and HBM4 line buildouts. Samsung Memory is comparable at 12-15% of revenue, with HBM4 PyeongTaek line build-out as the principal driver. Micron is approximately 10-12% (Boise / Idaho HBM ramp). TSMC and Intel together represent another 15-20% (logic etch + ALD demand for N3 / N2 / 18A). Kioxia / Western Digital is the 3D NAND etch demand layer. The customer concentration is similar to KLAC but with a higher memory tilt — which means the equity has more memory-cycle beta than KLAC.
The memory-cycle beta needs to be understood in detail. Memory-equipment revenue at Lam swung from approximately 30% of total in 2023 (the trough) to 50%+ in 2026 (the current peak). The principal cycle question is whether the 2026-27 memory order book holds or compresses. SK Hynix and Micron HBM capex is contracted under LTAs and effectively non-cyclical through 2027; Samsung HBM4 capex is more discretionary but tied to NVIDIA qualification timing rather than memory ASPs; Kioxia / WD 3D NAND capex is the most cyclical layer and is the most exposed to a NAND ASP correction. Aggregate exposure: roughly 60-70% of the memory book is HBM-linked (low cycle-beta) and 30-40% is NAND-linked (higher cycle-beta).
Competitive context
Applied Materials is the largest competitor across the broad equipment portfolio but only meaningfully competes with Lam in select etch and deposition niches. TEL (Tokyo Electron, 8035.T) is the principal direct competitor in high-aspect-ratio etch for 3D NAND and in select ALD applications — TEL has been gaining share at Kioxia and at the Chinese memory makers. ASM International is a competitor in ALD specifically. The structural advantage Lam holds is the installed-base service revenue (roughly 35% of revenue, growing 15% YoY) and the multi-decade customer co-development partnerships at the HBM lines.
The HBM-specific etch franchise is the deepest moat — Sense.i Plus has been refined across HBM2 / HBM3 / HBM3e generations with customer-specific recipes, and the install base at SK Hynix M16 and Samsung PyeongTaek is too embedded to be displaced by a TEL or AMAT tool on a single product transition.
Terminal risk
Two terminal risks. First, the same China-export tightening that affects KLAC — Lam has roughly 20-25% China revenue exposure (declining post-restrictions) and further tightening compresses that line. Second, the longer-tail risk that as 3D NAND layer counts saturate around 400+ layers, the etch-tool replenishment cycle slows — though the offset is that HBM stack heights are simultaneously rising (12-Hi to 16-Hi to 20-Hi) and TSV density is growing per die, which sustains Lam volume even if NAND moderates.
Bull case
The base case is FY26 revenue of $22-23B (versus FY25 around $17B), FY27 revenue of $25-27B, gross margin holding 48-50%, and non-GAAP EPS rising from FY26 around $5.80 toward FY27 $6.80-7.20. At 28x forward — discount to KLAC, broadly in line with the equipment cohort — that prints the stock at $190-200 post the well-flagged 10-for-1 split that came earlier in 2024, equivalent to current levels with modest upside.
The upside case is HBM4 + HBM5 capacity additions running well above current sell-side modelling — Trendforce now forecasts HBM unit growth of +60-80% in 2026 with another step in 2027 — which lifts Lam HBM-specific revenue by an incremental $1-1.5B annually through 2027. The 3D NAND layer-count ramp at Kioxia, Western Digital, and the Samsung NAND side adds another $400-700M. Combined, the upside case prints FY27 EPS toward $8.00 and the stock at $310-340 in 2027.
The HBM3e to HBM4 generational transition is the principal near-term driver. The HBM4 die ships in 12-Hi configuration mainstream and 16-Hi configuration premium, with each step requiring proportionally more TSV etches per die. The base-die migration from a DRAM process to a 5nm logic process at TSMC (an SK Hynix decision for HBM4 specifically) requires additional CD-SEM and metrology pull-through that benefits Lam's adjacent ALD and conductor-etch lines. The aggregate Lam revenue per HBM4 wafer is approximately 25-30% higher than the equivalent HBM3 wafer — a structural mix improvement that compounds the unit-volume growth.
Gap / bear case
Three concerns. First, +24% in 30 days has compressed the risk-reward; the stock pulled back 4.5% on May 15 on macro / yields concerns, which is a small reminder that even high-quality equipment names trade with cyclical equity beta. Second, the FY26 guidance has tightened to "growth on AI demand" rather than explicit quantification — consensus has moved up the EPS revisions but the company has been cautious about extrapolation, and any sequential softening in the memory order book (which is possible if SK Hynix or Samsung pulls in CapEx) would generate a multi-quarter air-gap. Third, the China-export-restriction overhang remains live in much the same way as KLAC — Lam has been the most affected of the broad equipment cohort over 2022-25, and continued restrictions through 2026-27 compress the China revenue line.
The fourth concern relates to share dynamics specifically. While Lam holds 80%+ share at HBM TSV etch, TEL has been investing in competing tools and has been gaining share at the 3D NAND high-aspect-ratio etch tier — Kioxia / Toshiba memory has been historically a TEL-leaning customer, and the Samsung NAND book is more competitively distributed than the HBM line. If TEL captures incremental NAND etch share through the 2027 generation transition, Lam's NAND-equipment revenue compresses against expectations.
Optionality
Three options. Hybrid-bonding integration alongside BESI / ASMPT — Lam has IP exposure here through the LRCX-BESI joint development on the HBM hybrid-bonding back-end. Continued China-domestic memory build-out (CXMT, YMTC) generating equipment demand at the tools that remain unrestricted. And the M&A optionality at the equipment-vendor consolidation level — both LRCX-AMAT and LRCX-BESI scenarios have been floated by sell-side over 18 months; while neither is base case, either becomes a one-time premium event.
The trade
LRCX — BUY 8/10. Entry: starter at $285 spot; size up on $260-270 pullback into the 50-day moving average. Position size: 2.5-3% of NLV as the etch/deposition equipment leg, paired against KLAC as the inspection / metrology leg. Stop: daily close below $245 (200-day MA breach and prior structural support). Catalyst date: July 22 Q4 FY26 print; September SEMICON West commentary on memory order book; any Trendforce or SemiAnalysis HBM capacity revision. Trim/exit triggers: memory-equipment order book softening below 25% YoY; gross margin guide below 47%; any verifiable China-restriction tightening that removes >$1B of annual revenue. Conviction: the cleanest HBM TSV etch monopoly and the second-best equipment franchise in the cycle after KLAC. Treat LRCX + KLAC as a pair — owning both at 5-6% combined gives you the equipment-cycle exposure without single-name concentration risk.