Investors Who Have Been Following AI Chips and Memory Chips May Be Forgetting About the Vital Industry That Enables Both of Them

AI-chip and memory demand is driving accelerating semiconductor-equipment spending, benefiting Applied Materials, Lam Research, and Cohu. Applied Materials reported 25% year-over-year fiscal Q3 sales growth and targets $10.25B in fiscal Q4 revenue, implying more than 12% sequential growth; Lam Research expects Q3 revenue of roughly $8.1B at the midpoint, up 21% sequentially. Cohu delivered 38% year-over-year Q2 revenue growth, expanded its AI-compute pipeline to about $850M, and narrowed its net loss to $0.2M from $16.9M a year earlier.

Analysis

The key investable distinction is that front-end wafer-fab equipment monetizes each process-node transition and rising memory-layer complexity, while AI-chip designers remain exposed to a narrower set of hyperscaler procurement decisions. AMAT and LRCX should therefore show better earnings durability than the market assumes if leading-edge logic and HBM capacity additions persist into 2027; service/installed-base revenue also cushions a later capex normalization. The second-order beneficiary is KLA (KLAC), whose inspection intensity rises with advanced packaging, HBM stacking, and yield complexity—an arguably cleaner way to express the same theme than chasing already-extended equipment momentum.

COHU is a higher-beta, lower-quality expression: test demand can rise with AI compute volumes, but its pipeline is not equivalent to booked revenue and its path to sustained profitability depends on utilization, customer qualification cycles, and operating-expense discipline. Unlike AMAT/LRCX, COHU has less pricing power and greater exposure to the broader analog/industrial recovery, so a weak non-AI semiconductor cycle could overwhelm the AI narrative. Treat margin conversion and order backlog—not opportunity-pipeline commentary—as the validation metrics over the next two quarters.

Near-term risk is that equipment stocks have already discounted a prolonged capex upcycle. A 1-3 month reset in hyperscaler AI spending, HBM pricing, or memory-maker capex plans would hit WFE names before reported revenue, while tighter China export restrictions could create an additional downside leg for the group. Over 6-18 months, the contrarian point is that equipment demand may remain firm even if AI accelerator unit growth moderates, because yield-improvement and packaging requirements increase process steps per useful chip; the thesis fails if WFE bookings roll over for two consecutive quarters or managements cut 2027 revenue outlooks.

Key Decisions for Investors

• Prefer a staged long AMAT/LRCX basket over NVDA on any 8-12% semiconductor-equipment pullback, with a 6-12 month horizon. Target relative outperformance versus SOXX if WFE bookings and service revenue remain positive; exit if either company guides the next quarter below consensus or cites a broad memory-capex deferral.
• Add KLAC as a quality relative-value leg: long KLAC versus short equal-dollar SOXX for 3-6 months. Inspection/metrology content should benefit from advanced-node yield challenges, while the hedge reduces broad AI multiple-risk; cover the short if SOXX underperforms KLAC by more than 15% or if export-control changes disproportionately impair KLAC China revenue.
• Do not initiate a core COHU long until the next earnings release demonstrates positive operating income or backlog/order momentum sufficient to support it. If confirmed, use a small 1-2% risk-budget position with a 9-12 month horizon; upside is material multiple expansion from sustainable profitability, but stop on renewed revenue-guide cuts or gross-margin deterioration.
• Monitor Samsung, SK Hynix, Micron, and TSMC capex commentary plus HBM contract pricing over the next two earnings cycles. A synchronized reduction in 2027 capex intentions is the actionable signal to cut equipment exposure, regardless of near-term revenue beats.