Applied Materials (AMAT.US) and Lam Research (LRCX.US) remain Bank of America's top selections in the semiconductor equipment sector, as the bank now anticipates that equipment expenditures will continue to grow. Following a volatile earnings season, the bank has updated its forecasts for wafer fabrication equipment. The expansion of AI wafer demand has led all major semiconductor companies to highlight an exceptionally strong market environment, with a promising outlook spanning 12 to 24 months. Companies like Lam Research (showing over 23% year-over-year growth), KLA Corporation (KLAC.US) (nearly 10% year-over-year growth), and Applied Materials (with a 20% year-over-year increase in systems business) have provided particularly impressive guidance for 2026, which is expected to accelerate growth into 2027, according to analyst Vivek Arya in a client note. Arya now projects chip equipment spending to reach $140 billion in 2026, $171 billion in 2027, and $193 billion in 2028, up from prior estimates of $131 billion, $150 billion, and $155 billion, respectively. He also forecasts spending levels of $186 billion for 2029 and $201 billion for 2030. Arya added that the industry expansion, with an expected compound annual growth rate of 18% from fiscal year 2025 to 2028, may still face supply constraints. However, semiconductor sales are projected to grow from approximately $780 billion in fiscal year 2025 to nearly $1.3 trillion by fiscal year 2028, reflecting a compound annual growth rate of 17% and indicating sustained healthy market momentum. The demand from agent AI is straining capacity for leading-edge logic, creating a need for additional production capacity. As XPU/accelerator roadmaps advance and agent AI gains prominence, competition for the scarce leading-edge logic capacity is expected to become exceptionally intense. Arya further analyzed that competition is fierce in the 3-nanometer process domain, where NVIDIA's (NVDA.US) Rubin chip will compete with AMD's (AMD.US) MI350X and MI400X, Google's (GOOGL.US) TPU v7, and Amazon's (AMZN.US) Trainium 3 chip. He explained that bottlenecks also exist in the CPU sector, as Intel (INTC.US) and AMD have raised prices, and NVIDIA's Vera CPU further intensifies market tightness. Switches have also achieved 3-nanometer-level processes, with NVIDIA's NVLink 6 and companies like Marvell Technology (MRVL.US) and Broadcom (AVGO.US) increasing their output capabilities. Arya pointed out that the scaling of AI chips for rack-scale architectures, the expansion of inference/agent workloads, and the diversity of design starting points are driving wafer demand far beyond current supply levels. Taiwan Semiconductor Manufacturing Company (TSM.US), which has reportedly pushed fab delivery timelines to 2027, and Intel/Samsung are racing to catch up. Furthermore, while it is premature to assess Elon Musk's Terafab project, its initial planned capacity of 100,000 wafers per month could potentially generate over $20 billion in fab equipment revenue in the long term. Arya stated that semiconductor equipment manufacturers could also benefit from sustained demand for memory, as AI continues to drive needs for high-bandwidth memory and DRAM. Additionally, Micron Technology's (MU.US) 25% year-over-year increase in capital expenditure for acquiring fabs to boost capacity further heightens the urgency. Arya added that while prices may stabilize, long-term supply agreements provide reliable visibility and give memory suppliers confidence in cash flow, enabling continued investment. The ability of suppliers to implement differentiated pricing and charge premiums for HBM and increasingly prevalent DRAM also serves as an incentive for their investment. The upgrade of NAND flash memory to 300 layers is expected to maintain strong momentum through 2027.
Comments