Beneath the grand claims and vast capital spending, AI is delivering weak productivity, mounting losses, and growing financial strain. If the boom cannot generate profits, jobs, or durable gains, how much longer can the bubble hold?
Four red flags signal that the U.S. economy has reached the peak of the AI bubble: (i) escalating liquidity shortages for key AI firms, unable to recoup rising capital and operational costs; (ii) growing competition and a building AI price war; (iii) a budding private credit crisis, related to its overexposure to software firms and AI data centers; and (iv) stagflation, caused by Trump’s tariffs and the Iran war and leading to higher energy costs and higher interest rates. However, even absent these headwinds, the AI boom cannot be sustained, because the aggregate net productivity impacts of the novel AI tools will be disappointingly small, will peter out rather quickly over time and will not be large enough to justify the stratospheric capital expenditures on AI models and AI infrastructure. The AI-driven ‘white-collar bloodbath’ will not happen, because even though AI tools will replace some tasks, these will also create new (often unproductive overhead) tasks, and augment and transform occupations, rather than completely destroy them. The productivity impacts of AI will disappoint, because AI tools do not just have positive effects on labor productivity, but also generate significant negative impacts — mostly because the use of AI creates new tasks and jobs, because the AI tools have to be monitored, supervised and curated, as the AI tools generate errors, slop, cybersecurity risks, brittle code and mounting technical debt that could kill businesses and institutions if it concerns mission-critical activities.
