Semiconductor equipment stocks: delayed beneficiaries of AI capital spending

The initial beneficiaries of the AI investment boom were GPU designers, followed by server, optical communication, liquid cooling, and data center power equipment vendors. In contrast, semiconductor equipment companies exhibit a slower earnings response—not because they are unrelated to AI, but because capital expenditure takes time to propagate down the supply chain: cloud companies first increase their compute budget, chip designers secure orders, and then foundries, memory, and packaging fabs decide on capacity expansions, finally placing orders with equipment vendors for lithography, etch, deposition, metrology, and advanced packaging tools.

This lag is precisely why equipment stocks are referred to as 'second-phase beneficiaries' of AI-related capital spending. When GPU supply is tight, the market focuses most on per-chip pricing; once demand is confirmed as a multi-year trend, attention shifts from chip inventory levels to insufficient manufacturing capacity. Taiwan Semiconductor plans approximately $40.9 billion in capital expenditures for 2025, with its 2026 budget further increasing to $52–56 billion, about 70–80% of which will be allocated to advanced nodes. This isn't merely chasing short-term orders—it's strategic preparation for 2nm, A16, advanced packaging, and overseas capacity.Taiwan Semiconductor data

From buying GPUs to purchasing the 'tools that make GPUs'

Equipment demand is not only driven by logic chips. AI servers require large amounts of HBM, and with each new generation of HBM, the number of stacked layers, through-silicon vias (TSVs), and bonding steps increases—raising the utilization intensity of etching, thin-film deposition, cleaning, inspection, and packaging equipment. In other words, even if wafer volume growth remains limited, the number of process steps per wafer could still rise, thereby expanding the addressable market for equipment vendors.

Each beneficiary has its own strengths. ASML holds a critical position in EUV lithography; Applied Materials covers deposition, ion implantation, and advanced packaging; Lam Research excels in etch and deposition equipment; and KLA benefits as increasingly complex processes make yield control ever more crucial. Tokyo Electron expects sales of new equipment in the first half of its fiscal year ending March 2027 to grow 41% year-over-year, and forecasts advanced packaging revenue to surge by over 60%, driven precisely by accelerated investments in high-end logic, DRAM, and HBM.Tokyo Electron data

Recent results already show this trend is materializing. Applied Materials reported Q2 FY2026 revenue (ending April 2026) of $7.91 billion, up 11% year-over-year, and expects its semiconductor equipment business to grow by more than 30% for the full year. Lam Research reported Q3 FY2026 (ending March) revenue of $5.84 billion, up 9% quarter-over-quarter, with an operating margin of 35%. These figures indicate that AI investment is gradually shifting from chip design companies down to the manufacturing equipment layer.Applied MaterialsandLam Research

However, 'delayed benefit' does not mean immunity to cyclical risks. Wafer fab equipment orders involve substantial sums; if AI demand falls short of expectations, customers delay fab construction, or HBM supply shifts from shortage to oversupply, equipment orders can be quickly postponed. Equipment makers also face high customer concentration, and revenue recognition is subject to timing of delivery, installation, and acceptance. Moreover, exposure to the Chinese market, export controls, and tariff policies could alter product mix and profitability.

Investors therefore should look beyond AI hype and instead monitor wafer fab capital expenditures, equipment orders, order backlogs, customer prepayments, and service revenue. Service and spare parts revenue is typically more stable than new tool sales and can buffer against cyclical volatility. Conversely, if stock prices have already priced in several years of capacity expansion, even realized earnings growth may not be sufficient to drive further valuation upside.

The first wave of the AI industry rally focused on who could design the fastest chips; the second wave asks who can deliver sufficient manufacturing capacity. As capital spending shifts from buying GPUs to purchasing the 'tools that make GPUs and HBM,' equipment stocks are stepping out from behind the scenes. Their breakout comes later and carries heavier cyclicality, yet they may represent the group of 'shovel sellers' with the deepest technological moats and highest order visibility in the AI infrastructure boom.

(Chip & Compute Series No. 76)

$NVIDIA (NVDA.US)$$Advanced Micro Devices (AMD.US)$ 、 $Intel (INTC.US)$ 、 $Intel (INTC.US)$ 、 $Marvell Technology (MRVL.US)$ 、 $Qualcomm (QCOM.US)$ 、 $Arm Holdings (ARM.US)$

161K Views