// ai economic impact

Source: Semianalysis

The productivity paradox that haunted early computing—where massive IT investment produced no measurable economic gains—is repeating with AI, except now we're deploying vastly larger models without visibility into what they actually produce. Unlike software that generates discrete, countable outputs, most frontier AI deployment happens in corporate black boxes where the work (research synthesis, code generation, decision support) either never touches recorded economic activity or gets absorbed into existing line items, making the real multiplication effect invisible to traditional metrics. This measurement gap matters because policymakers, investors, and boards are making trillion-dollar infrastructure bets on faith rather than data about whether these systems are creating genuine productivity gains or just automating expensive white-collar work that was already accounted for.