Industry Analysis
Citi’s renewed focus on AI compute semiconductors is less about demand optimism and more a bet on the scarcity of sub-3nm manufacturing capacity. Only TSMC (Taiwan, China) currently delivers high-volume EUV-based 3nm production, effectively locking in NVIDIA’s next-gen AI chip orders for 2026–2027. This dominance triggers ripple effects: EDA vendors must accelerate 3nm design rule support, while OSATs scramble to scale CoWoS packaging for chiplet-based architectures. Yet U.S. export controls compel TSMC to diversify fabs to Arizona and Japan, inflating capex and supply chain fragility. Samsung’s GAA transistor push remains hamstrung by yield issues, cementing TSMC’s lead. Over the next 12–24 months, advanced packaging—not just lithography—will define competitive advantage. Investors should hedge against overreliance on a single node by prioritizing chip designers with multi-foundry portability.
This page displays AI-generated summaries and metadata for research purposes. Original content belongs to the respective publishers.