Applied Materials CEO says agentic AI drives wafer demand

Agentic AI is not a compute-scale story—it is a workload-architecture shift. Inference migrates from bursty batch jobs to sustained, always-on, high-throughput processing. That converts wafer demand from cyclical spikes into a structural plateau, and Applied Materials' 20%+ revenue trajectory is the sharpest leading indicator of that regime change. The true bottleneck is yield, not capacity. Below 2nm, a single percentage-point yield gain at a leading foundry saves billions in capex. Etch, ALD, and CMP—core AMAT competencies—sit at the exact chokepoint where yield is won or lost. The CEO's emphasis on yield over raw volume is the most consequential signal in this cycle. Competitively, Lam Research will escalate its advanced-etch share fight while KLA tightens its inspection-and-metrology moat. AMAT must complete its GAA-generation platform transition by 2026 or cede structural ground. A subtler risk: domestic Chinese equipment makers are eroding the 28nm-and-above profit pool, while export controls simultaneously cap advanced-node shipments to China—squeezing AMAT's margin base from both ends. The 12-to-24-month tail: HBM4 stacking, 2nm logic, and CoWoS advanced packaging create a triple wafer-start overlay. Agentic AI is a usage-paradigm migration, not a product cycle. Once the demand curve flattens into a plateau, equipment vendors' pricing power shifts structurally upward—and that repricing has not yet fully hit the stock.