General-purpose technologies
A general-purpose technology is a widely applicable input with continuing room for improvement and strong complementarities with innovation in the sectors that use it.
Explanation
Steam power, electricity, and computers did not transform the economy through one finished invention. They became useful across many activities, kept getting better, and encouraged businesses to invent new machines, products, and ways of organizing work around them.
Those complements take time and are produced by different actors. A technology supplier may underinvest when much of the benefit goes to users, while users may wait because they are unsure how fast the core technology will improve. The result can be broad potential alongside slow adoption and delayed productivity gains.
AI has several possible general-purpose features, but the label is a hypothesis rather than a measured conclusion. Models must still become dependable products, and firms must redesign workflows, data systems, skills, and institutions before model capability becomes economy-wide output.
Examples
-
Electrifying a factory
Replacing a steam engine with an electric motor produced limited gains when a factory kept the old layout. Larger gains arrived when firms reorganized production around smaller motors and flexible assembly lines.
-
Adopting AI
Access to a capable model may help a few workers immediately, but broad gains can require new software, evaluation systems, permissions, training, and a different division of work.
Related timeline entries
- The AI expectations gap (16 October 2026)
- China is more exposed to semiconductor supply-chain disruptions than the US (6 October 2026)
- Europe 2031: What getting AI wrong means for us (11 June 2026)
- Forecasts of AI & Economic Growth (6 November 2025)
- Economics and Transformative AI (2 October 2025)
- AI as Normal Technology (15 April 2025)
- The Productivity J-Curve: How Intangibles Complement General Purpose Technologies (January 2021)
- General Purpose Technologies: Engines of Growth? (January 1995)