Source: Substack
The consensus narrative around AI as an economic accelerant relies on historical comparisons to electricity that don't hold. Electricity spread universally across sectors within decades. AI adoption faces genuine bottlenecks: integration costs, skill gaps, and the difficulty of deploying models in legacy systems that still dominate advanced economies. The 10-15 year GDP forecasts circulating among venture capitalists and some economists mistake productivity potential for actual deployment velocity. They confuse what's theoretically possible with what companies will actually implement given capital constraints, regulatory friction, and the long replacement cycles of existing infrastructure.