Wall Street Is Misreading the AI Chip Selloff - Dr. Robert Castellano's Semiconductor Deep Dive Newsletter

The recent semiconductor selloff has been interpreted by Wall Street as a sign of slowing AI spending, but Dr. Robert Castellano's latest deep-dive analysis reveals a different story. Contrary to mark... Read original →

Wall Street’s panic over the AI chip selloff reveals a fundamental misreading of semiconductor dynamics. HBM shipments are set to surge over 30-fold by 2027—not a sign of waning demand, but of infrastructure scaling. This memory bottleneck directly fuels NVIDIA’s Blackwell and AMD’s Instinct, while straining TSMC’s CoWoS capacity. Technologically, soaring EUV and 3nm wafer costs are accelerating hyperscaler in-house ASIC development (e.g., Google’s TPU, Amazon’s Trainium). Geopolitically, U.S. export controls inflate supply chain redundancy costs, yet Taiwan, China-based foundries remain irreplaceable due to packaging dominance. In response, Dell and Vertiv are bundling HBM-optimized liquid cooling to lock in data center deals. Over the next 12–24 months, HBM will serve as the leading indicator of AI capex health, while GaN players like Innoscience benefit from the energy-efficiency tailwinds of next-gen AI data centers.