The Installation–Deployment Cycle
Carlota Perez's 2002 framework, Technological Revolutions and Financial Capital, argues every technological revolution since 1771 has run the same two-act script: an Installation phase where financial capital funds infrastructure faster than the real economy can absorb it, a Turning Point where the resulting bubble corrects, and a Deployment phase where production capital takes over and the technology diffuses broadly. The live question for 2026 is where the AI buildout sits on that arc.
The Perez Framework
Perez's central claim is that a technological revolution is not one continuous story but two, separated by a crash. In the Installation phase, financial capital — equity markets, venture funding, credit — races to build the new infrastructure, and valuations decouple from earnings because capital is pricing a future that has not arrived yet. The Turning Point is the recession or crash that follows: overbuilt capacity meets a funding retrenchment, weaker players are cleared out, and regulation typically tightens. Only after that reset does the Deployment phase begin — production capital, not financial capital, takes over, and the technology's payoff shifts from building it to using it broadly across the wider economy, in what Perez calls the golden age.
Five Revolutions, One Script
Perez traces this pattern across five technological revolutions dating back to the first, beginning in 1771. Each one, in her account, repeats the same choreography: infrastructure investment running ahead of demonstrated demand, a frenzy of speculative capital chasing the theme, a crash that separates durable capacity from overbuilt excess, and then a longer deployment period in which the surviving infrastructure gets monetized across ordinary industries rather than just by the builders themselves. The book's core argument is that the crash is not evidence the technology was a mistake — it is a scheduled phase between building the infrastructure and profiting from it at scale.
Where the AI Buildout Sits
Applied to 2026, the framework gives testable rather than impressionistic criteria. Installation-phase signatures include a capex race funded well ahead of demonstrated revenue, valuation dispersion collapsing across anything AI-adjacent, and financing flowing to builders of infrastructure — compute, power, packaging — rather than to its users. Deployment-phase signals look different: productivity gains showing up broadly across industries that use AI rather than build it, profit growth migrating from infrastructure providers to appliers, and financing discipline returning as production capital replaces speculative capital. The honest read is that both signatures are visible at once in different parts of the stack — which is itself consistent with Perez's model, since the transition between phases is never a single clean date.