The Semiconductor Cost Curve: Why Chips Get Cheaper… Until They Don’t

For decades, learning curves and Moore’s Law pushed cost per transistor down. But at advanced nodes, the curve bends: EUV tools, mask sets, and yield challenges make fixed costs explode. Foundries respond with capacity pre-buys and long-term agreements to de-risk utilization, shifting risk back to customers. The result is a barbell world: ultra-high-volume designs justify bleeding-edge nodes; everyone else stays on mature processes where total cost of ownership wins. For hardware startups, the economic sweet spot is often clever system design on established nodes, plus firmware differentiation—not chasing the shiniest geometry.