AI investment bubble vs historical infrastructure cycles

🔑 Enhanced Key Takeaways

• •The 'deployment age' concept, popularized by economist Carlota Perez, suggests that the current AI phase is transitioning from the 'installation period' (characterized by financial bubbles) to a 'deployment period' where real-world productivity gains materialize.
• •Recent data indicates that while AI capital expenditure (CapEx) by hyperscalers has reached record highs, the 'revenue gap'—the discrepancy between infrastructure spend and immediate AI-driven software revenue—is widening compared to the 1990s internet boom.
• •Energy constraints have emerged as a unique bottleneck in the current AI cycle, with power grid capacity and data center cooling requirements acting as physical limits that did not constrain the software-centric telecom boom of the early 2000s.
• •Institutional investors are increasingly shifting focus from 'model performance' metrics (like MMLU scores) to 'unit economics' and 'inference cost per token,' signaling a maturation of the investment thesis toward profitability.
• •Historical analysis of the 1840s 'Railway Mania' shows that while most railway companies went bankrupt, the underlying infrastructure created a permanent reduction in transportation costs that fueled the Industrial Revolution, mirroring the potential for AI to permanently lower the cost of cognitive labor.

🔮 Future ImplicationsAI analysis grounded in cited sources

⏳ Timeline