Lam Research: Etch and Deposition Tools in the AI Wafer Expansion Cycle

Executive Summary

Lam Research (LRCX) is the global leader in etch and deposition equipment for semiconductor manufacturing, generating $14.9 billion in fiscal 2023 (ended June 2023). While smaller than Applied Materials by revenue, Lam holds dominant market positions in specific critical process steps — particularly atomic layer etch, plasma etch for NAND flash, and low-pressure CVD deposition — that are essential to manufacturing both AI logic chips and the memory required to make AI systems function. The AI buildout requires not only more GPU and CPU manufacturing capacity but also dramatically more DRAM and high-bandwidth memory (HBM) capacity, where Lam is particularly well-positioned. This report concludes that Lam Research faces minimal AI-driven margin compression risk and benefits from a structural tailwind as AI drives memory and advanced logic capacity expansion simultaneously.

Business Through an AI Lens

Lam Research's product portfolio is concentrated in three process categories: etch (plasma and atomic layer etch for pattern definition), deposition (chemical vapor deposition, atomic layer deposition, and physical vapor deposition for film formation), and clean (post-process residue removal). These are foundational manufacturing steps that every wafer must undergo multiple times during fabrication.

The AI connection to Lam is direct and quantifiable through memory. AI training and inference requires massive amounts of DRAM — an NVIDIA H100 SXM5 module contains 80GB of HBM3, and a typical eight-GPU server contains 640GB of stacked DRAM dies. Each HBM stack consists of 8-12 DRAM dies with thousands of through-silicon vias (TSVs) — and TSV formation requires etch and deposition steps where Lam is the market leader. As AI servers proliferate globally, HBM demand is growing at rates that may exceed 200% annually, creating an exceptional demand environment for Lam's memory equipment.

Beyond HBM, NAND flash memory — used for AI training data storage and model checkpoint storage — is undergoing its own capacity expansion. 3D NAND manufacturing at 200+ layer stacks requires extremely high-aspect-ratio etch (HARE) processes that Lam dominates. Every layer added to a NAND stack increases Lam's equipment content per wafer.

For logic manufacturing (AI GPUs and CPUs), Lam is important but less dominant than in memory. TSMC's advanced node process flows require competitive etch and deposition equipment, and Lam competes with Tokyo Electron (TEL) and Applied Materials across multiple process steps. Lam's atomic layer etch products, which enable nanometer-scale patterning with atomic precision, are particularly critical for sub-3nm logic.

Revenue Exposure

Lam's revenue by end market demonstrates the critical importance of memory to its business model.

The memory market's 2023 downturn — driven by the collapse of PC, smartphone, and server DRAM demand as the world worked through post-COVID inventory excess — reduced Lam's revenue significantly from the $17.5 billion peak in fiscal 2022 to $14.9 billion in fiscal 2023. Memory customers dramatically cut equipment purchases as chip prices fell 60-70% in some categories.

However, AI is changing the memory demand equation fundamentally. HBM is an AI-specific product that commands pricing 3-5x conventional DRAM. SK Hynix, Samsung, and Micron are all investing heavily in HBM capacity, which is Lam's highest content-per-wafer product. NAND demand for AI data centers is also recovering as the hyperscalers rebuild storage capacity to serve AI workloads.

China revenue, similar to Applied Materials, represents a meaningful exposure — approximately 25-30% of Lam's revenue in recent years. Export control restrictions create uncertainty, but Lam's mature-node memory equipment (for legacy NAND and DRAM production) has so far been less restricted than leading-edge logic equipment.

Cost Exposure

Lam Research's cost structure resembles Applied Materials but with higher revenue concentration in memory equipment, which historically has lower average selling prices but higher volume than logic equipment. Gross margins have ranged from 43-47% over the past five years, below Applied Materials' 47-49%, reflecting the more price-competitive memory equipment market.

R&D spending of approximately $1.7 billion in fiscal 2023 (approximately 11% of revenue) focuses on next-generation etch and deposition processes for sub-3nm logic and 300+ layer NAND. The critical R&D challenge is maintaining process leadership as chipmakers demand atomically precise results across increasingly complex 3D device structures.

Input cost inflation in precision components — vacuum systems, ceramic components, plasma generators — has been modest but persistent. Lam sources globally and has diversified supply chains, which partially insulates it from single-supplier price shocks. AI does not directly inflate Lam's input costs meaningfully.

The primary cost risk is talent competition. Plasma physics, materials science, and process engineering expertise is the core of Lam's competitive advantage, and demand for these skill sets is growing as chipmakers and equipment competitors expand. Lam's Fremont, California headquarters puts it in competition with Silicon Valley technology companies for engineering talent, creating upward wage pressure.

Moat Test

Lam's moat in memory equipment is the strongest in its product portfolio. The company's process recipes — the specific sequences of gas chemistries, pressures, temperatures, and plasma conditions that reliably etch or deposit films — are the result of decades of co-development with SK Hynix, Samsung, and Micron. These recipes are integrated so deeply into chipmakers' manufacturing processes that switching equipment vendors requires a full re-qualification effort typically lasting 18-24 months and costing tens of millions of dollars.

In HARE etch for 3D NAND, Lam holds estimated market shares above 75%, creating near-monopoly economics in this critical process step. No competitor has been able to match Lam's combination of etch rate uniformity, selectivity, and profile control at aspect ratios above 60:1.

The logic market is more competitive. TEL and Applied Materials are strong competitors in CVD deposition and certain etch applications. However, Lam's atomic layer etch (ALE) technology is differentiated for sub-3nm features and is gaining adoption at TSMC and Samsung as they advance to N2 and GAA architectures.

Timeline Scenarios

1-3 Years (Near Term)

Memory capex recovery drives Lam revenue back toward $18-20 billion by fiscal 2025-2026 as HBM capacity investments accelerate and NAND supply/demand rebalances. Gross margins recover to 46-48% as revenue scale absorbs fixed R&D and overhead costs. Advanced packaging (particularly TSV etch for HBM) grows from a modest contributor to a $1.5-2.5 billion annual revenue line.

3-7 Years (Medium Term)

3D NAND stacks reach 500+ layers, with each additional 100 layers requiring proportionally more HARE etch steps and Lam equipment. DRAM scaling transitions to 3D DRAM architectures, which may require new etch and deposition techniques. AI inference at the edge creates demand for embedded memory in edge AI chips, further expanding the addressable market. Lam revenue trajectory targets $22-26 billion in this window.

7+ Years (Long Term)

The fundamental question is whether novel memory architectures — resistive RAM, spin-transfer torque MRAM, phase-change memory — displace NAND and DRAM at scale. These alternatives have different manufacturing process requirements, and Lam must invest in new process capabilities to remain relevant. The transition timescale (10-15 years) provides sufficient runway for adaptation.

Bull Case

In the bull case, HBM demand grows faster than SK Hynix, Samsung, and Micron can add capacity, driving sustained equipment purchases through 2028. 3D NAND layer counts accelerate to 400+ layers by 2027, expanding Lam's HARE etch content per wafer by 40-50%. Advanced packaging for chiplet AI processors becomes a $4-5 billion annual Lam revenue opportunity. Total revenue reaches $26-30 billion, gross margins expand to 50%+, and Lam trades at 20-25x forward earnings.

Bear Case

In the bear case, AI chip demand plateaus in 2026, memory capex enters another correction cycle, and export controls eliminate China revenue entirely. NAND layer count growth slows as chipmakers prioritize yield over density. Lam revenue falls to $12-14 billion in the trough, gross margins compress to 40-42%, and the stock de-rates to 12-15x earnings. However, the HBM structural demand driven by AI makes a severe trough less likely in this cycle than in prior memory cycles.

Verdict: AI Margin Pressure Score 2/10

Lam Research earns a 2 out of 10 — among the most protected large-cap semiconductor companies in the AI transition. Memory equipment is the picks-and-shovels business of AI infrastructure: every AI server requires HBM, and every HBM die requires Lam equipment to manufacture. The business faces cyclical risk (memory capex cycles) and geopolitical risk (China export controls) but not structural AI disruption risk. Lam's near-monopoly in HARE etch for 3D NAND and its dominant position in HBM TSV etch represent durable competitive advantages that AI investment only strengthens.

Takeaways for Investors

Lam Research is a core beneficiary of the AI-driven memory capacity expansion cycle. Investors should track HBM capital expenditure announcements from SK Hynix, Samsung, and Micron as leading indicators of Lam's equipment order trajectory, which typically leads revenue recognition by 6-12 months. NAND layer count progression is a valuable long-term demand proxy — every 100-layer increment in industry-leading NAND stack height adds approximately $1-2 billion to Lam's addressable market. China export control news flow is the primary near-term risk management event to monitor. Lam's history of significant cash return — buybacks and dividends — provides a total return cushion during equipment cycle troughs, making it a quality cyclical holding for patient capital.