AMD CEO Lisa Su says the company plans to “substantially increase” its supply in 2027 and needs “more advanced wafer capacity”, as she meets Foxconn and TSMC
Context & Ripple Effects
AMD’s AI-chip business had already moved from an early MI300 supply constraint—MI300 availability was expected to stay tight through 2025—to a larger push for AI revenue, which AMD said reached $5 billion in 2024. The 2027 supply target turns that demand trajectory into a manufacturing-capacity requirement.
TSMC had separately said its $100 billion US commitment would expand Arizona capacity amid a multiyear demand trend. Su’s meetings with TSMC and Foxconn place AMD’s volume plans within that broader competition for advanced-chip production.
First-order effects
- AMD must secure additional advanced-wafer capacity to support the supply increase it has set out for 2027, making TSMC allocation a central operating constraint.
- TSMC receives another explicit demand signal from a major chip customer, while Foxconn is part of AMD’s discussions around scaling the supply chain.
Second-order effects
- AMD’s ability to convert AI-chip demand into 2027 shipments is increasingly governed by foundry capacity rather than chip demand alone, extending the supply pressure seen with MI300.
- TSMC’s capacity expansion becomes more consequential for customers such as AMD because added output must be allocated among multiple large buyers seeking advanced wafers.
Third-order effects
- If this pattern persists, access to contracted advanced-node capacity will become a more important competitive differentiator for AI-chip vendors than design execution alone.
- AI infrastructure growth is pushing supply-chain planning further upstream, with chip designers coordinating earlier with foundries and manufacturing partners to secure volume.
The trend: AI-chip competition is becoming a capacity-allocation contest in which advanced-wafer supply sets the pace of product availability.