At TSMC's June 4 annual shareholders meeting, CEO C.C. Wei announced that chip supply will fall short of AI-driven demand for years. Advanced nodes are sold out through at least 2027, with demand exceeding capacity by an estimated 25–30% in 2026. Wei stated plainly: 'It will be a long time before we can meet customer demand.' The supply crunch is structural, not cyclical, and persists despite TSMC ramping 3nm capacity and building $165B in U.S. fab capacity.
TSMC reaffirmed 30%+ revenue growth for 2026 in USD terms and raised capex guidance to the high end of its $52–56B range. Q1 2026 revenue grew ~35% year-over-year. In response to tight capacity, TSMC is raising employee bonuses by ~30% (third consecutive year) and accelerating clean-room buildouts and tool procurement. Even with Arizona facilities ramping, the company cannot promise to satisfy American customers' full demand on any near-term timeline.
For architects, this signals a 3–4 year window of persistent chip scarcity that will constrain hyperscaler capacity additions and favor vendors who can secure long-term supply agreements. Multi-node diversity and alternate foundries (Samsung, Intel 18A) are no longer optional—they are core supply-chain resilience. Every month of delay in fab expansion now translates directly into delayed AI infrastructure deployments and higher costs.