Question · Q3 2025
Bruce Lu referenced a customer's forecast of a $3-4 trillion AI infrastructure opportunity by 2030, implying a 40% CAGR similar to TSMC's AI growth guidance. He asked for TSMC's specific forecast for AI infrastructure growth and token growth over the next five years, and if TSMC's AI-related revenue growth would track or even exceed the CapEx growth of major cloud service providers given the value TSMC captures. He followed up, pointing out that the exponential token growth rate appears substantially higher than TSMC's AI-related revenue CAGR, suggesting a widening gap over time, and asked how TSMC addresses this discrepancy and if it's considered a major issue.
Answer
CEO Che-Chia Wei confirmed that the mid-40% CAGR for AI demand aligns with major customers' forecasts, not including all infrastructure build-up. He stated that token growth is exponential, increasing almost every three months, which reinforces the demand for leading-edge semiconductors. He reiterated that TSMC is working hard to narrow the demand-supply gap. Che-Chia Wei explained that the discrepancy between token growth and TSMC's AI revenue CAGR is resolved by continuous technological improvement and node migration. As customers move to more advanced nodes, they can handle significantly more tokens more efficiently, and TSMC's technology combined with customer design changes can meet the exponential growth in token consumption.