RBC Capital Markets expects semiconductor revenue stemming from artificial intelligence applications to grow from $220B in 2025 to more than $550B by 2028.
“Tight supply is leading to extended visibility with managements highlighting 18-month backlogs,” said RBC analysts Srini Pajjuri and Grant Li in an investor report. “Infrastructure constraints could delay some projects but aren’t necessarily a negative in our view, as they may ultimately extend and smooth the AI spending cycle. We see limited threat to GPU dominance despite the recent custom ASIC progress given the rapid pace of AI technology evolution and long ASIC design cycles.”
This backdrop prompted the financial firm to initiate coverage on an array of semiconductor companies with an Outperform rating. This included Nvidia (NVDA), Micron (MU), Marvell (MRVL), Arm (ARM), Astera Labs (ALAB), ASML (ASML), Applied Materials (AMAT), Lam Research (LRCX) and Lattice Semiconductor (LSCC).
RBC applied Sector Perform ratings to Broadcom (AVGO), AMD (AMD), Intel (INTC), KLA (KLAC), Sandisk (SNDK), Qualcomm (QCOM), Skyworks Solutions (SWKS) and Silicon Laboratories (SLAB).
High-bandwidth memory demand is expected to be a key driver and should reduce the cyclical nature of the memory market.
“AI workloads are shifting to reinforced learning and distributed inferencing, which are Memory intensive,” Pajjuri noted. “Upcoming HBM4 transition is another material tailwind given 30-50% ASP premium. GenAI is also driving strong demand for high-capacity server DIMMs and NAND eSSDs. While high Memory prices could have an impact on PC/Smartphone demand, we expect Memory demand to outpace supply into 2027.”
Like several other financial firms have indicated, wafer fabrication equipment spending is also expected to be strong over the next two years.
“In addition, technology trends such as back-side power, advanced packaging, and 3D structures give us conviction that WFE can grow at least at a high single-digit rate for the next 2 years,” Pajjuri added.