Power Diffusion

In the AI Story

Nye introduced power diffusion as a defining feature of the information age. Previous technological waves — the printing press, radio, television — expanded individual and non-state capability but preserved the fundamental asymmetry between institutions and individuals. Building a printing press required resources. Operating a radio station required capital. Even the internet, which promised radical individual empowerment, ultimately consolidated power in platform companies whose influence exceeded many sovereign states. Communication tools enabled coordination; they did not enable individuals to compete with institutions on production.

AI disrupts this asymmetry. The individual with access to Claude Code is not merely able to talk to more people or organize more efficiently; she is able to build — to produce software, products, services that compete directly with organizational outputs. The capacity to build is the capacity to create durable economic value, and durable economic value is the foundation of durable power. This is why AI diffusion differs from earlier empowerment waves. The power that diffuses is not the power to speak but the power to make, and the power to make, in economic systems, is the power that compounds.

The geopolitical implications become visible when mapped onto global human capital distribution. Forty-seven million developers worldwide, with fastest growth in Africa, South Asia, and Latin America, represent latent capability that institutional barriers previously suppressed. The barriers were real but not intrinsic — they were artifacts of the cost structure of software development: years of specialized training, expensive infrastructure, proximity to ecosystems that turned code into products. When AI collapses the cost structure, the barriers fall and the latent capability activates. The resulting diffusion produces not a redistribution among states but a structural change in what states must govern and how.

The strategic response requires sophistication most national strategies lack. The prevailing frameworks — the U.S.-China AI race, European regulatory models, various national AI strategies — all operate on the assumption that AI power is something to be accumulated, defended, and deployed. Nye's framework suggests a different orientation: AI power must be shared strategically, governed wisely, and deployed in ways generating voluntary cooperation of others. A nation that hoards AI capability through export controls or restrictive licensing may preserve short-term hard power advantage but forfeits the soft power of being perceived as a generous and trustworthy partner. In a domain where capability diffuses through open-source models, academic publication, and sheer portability of AI knowledge, the hard-power advantage of hoarding is time-limited while the soft-power cost is permanent.

Origin

Nye developed the concept in The Future of Power (2011), building on earlier work with Robert Keohane on complex interdependence. The AI-specific extension recognizes that diffusion now operates through production capability rather than merely communication capability, a shift Nye anticipated in structure if not in specifics.

Key Ideas

Diffusion versus transition. Power diffusion moves capability between levels of organization; power transition moves it between states; the two phenomena require different strategic responses.

Production not communication. Earlier empowerment waves provided communication tools; AI provides production tools, enabling individuals to create durable economic value that compounds into durable power.

No counterparty. Diffusion cannot be managed through traditional diplomacy because there is no single counterparty with whom to negotiate; it must be managed through institutional design.

Latent capability activation. Global distribution of human capital means AI diffusion activates suppressed latent capability most rapidly in regions institutional barriers had most severely constrained.

Hoarding paradox. Restricting AI capability preserves short-term hard power but forfeits long-term soft power in a domain where diffusion is rapid and alternatives emerge quickly.

Debates & Critiques

Some analysts argue Nye overstates diffusion by understating the continued concentration of foundational AI capability in a small number of companies and states. The book treats this as accurate about the foundation layer and incomplete about the application layer: foundations concentrate while applications diffuse, and the applications are where production — and therefore durable power — actually occurs.