The AI Bubble Is Widely Misunderstood | Steve Hou

Key Concepts

• AI Bubble: A period of intense capital investment in AI-related infrastructure, characterized by high valuations and rapid technological adoption.
• Agentic AI: AI systems capable of autonomous decision-making and recursive task execution (e.g., AI calling other AI models), significantly increasing compute demand.
• Baumol’s Cost Disease: The phenomenon where sectors with low productivity growth (e.g., healthcare, childcare, manual trades) experience rising costs because they cannot easily automate, despite technological progress in other sectors.
• Jevons Paradox: The observation that as technology increases the efficiency with which a resource is used, the total consumption of that resource increases rather than decreases.
• Capital Expenditure (CapEx) Cycle: The massive investment phase in data centers, chips, and energy infrastructure driving current economic growth.
• Token Efficiency: The emerging need for businesses to optimize AI usage as "all-you-can-eat" subscription models give way to cost-reflective pricing.

1. The AI Investment Thesis

Steve Hoe argues that the AI boom is a legitimate technological revolution rather than mere hype. Unlike the dot-com bubble, which suffered from significant "unused capacity," the AI bubble is characterized by immediate, widespread adoption and a "wonkish" focus on physical infrastructure.

• The Chain Reaction: Hoe describes the AI cycle as a series of "nukes" (catalysts) that amplify each other. Initial skepticism was countered by the emergence of agentic AI, which shifted demand from simple chatbot queries to complex, recursive compute tasks.
• Underestimation: The market underestimated the recent acceleration because most observers do not code. Coders understand that agentic AI creates a non-linear, "hundred-fold" increase in compute demand, a reality often missed by non-technical analysts.

2. Macroeconomic Impact and GDP

AI investment has acted as a critical cushion for the U.S. economy, offsetting the cooling effects of the 2022 interest rate hikes.

• Direct Contribution: While consumption remains the primary driver of GDP, non-residential construction (data centers) has become a significant contributor to growth.
• Wealth Effects: The financialization of the economy means that the AI-driven stock market boom has created wealth effects, supporting consumption in the "top half" of the economy.
• Global Supply Chain: Countries like South Korea and Taiwan are seeing massive economic growth due to their role in the chip and memory supply chain, mirroring the U.S. investment cycle.

3. The Productivity Debate

There is currently no "cleanly identified" causal link between AI and aggregate labor productivity in official data.

• Measurement Challenges: Hoe notes that current productivity prints may be skewed by "composition bias" (shrinking labor-intensive sectors) and the "no-hire, no-fire" environment where existing workers are simply doing more with the same resources.
• Anecdotal Evidence: Citing Jeff Bezos, Hoe suggests that when data and anecdotes disagree, anecdotes are often right. Many firms report 20% efficiency gains, which will eventually manifest in broader economic data as the "time to build" and learning curves are overcome.

4. Monetary Policy and Fiscal Implications

• Preemptive Rate Cuts: Hoe is skeptical of justifying rate cuts based on the "unproven anticipation" of AI-driven disinflation. He argues that the Fed should focus on observable data, such as wage growth and services inflation, rather than speculating on AI’s future impact.
• The Debt Problem: The U.S. fiscal path is an "arithmetic problem" where tax receipts (stuck at 17–20% of GDP) cannot cover rising interest costs and primary deficits.
• The Growth Solution: The only viable path to fiscal sustainability is growing the denominator (GDP). However, Hoe warns that AI investment might inadvertently raise interest rates by increasing competition for capital, and that the "cornucopia" of free, limitless AI-driven abundance is still a distant, sci-fi prospect.

5. The Future of Economics

Hoe believes AI will fundamentally reshape economics by:

• Agentic Simulations: Allowing economists to run complex, realistic simulations of policy impacts (e.g., Fed balance sheet reduction) using autonomous agents.
• Tool Accessibility: Enabling researchers to utilize powerful, complex mathematical tools (like Kalman filters) with greater confidence and speed.
• Communication: Improving the ability of policymakers to explain complex economic outcomes to the public, potentially reducing the communication failures seen in recent years.

Synthesis

The AI boom is a sustained, infrastructure-heavy investment cycle that is currently in a "buildout" phase. While it is undeniably a bubble, its impact is real and distinct from previous tech cycles due to the immediate utility of the technology and the emergence of agentic AI. The primary economic risks involve the "Baumol’s disease" of physical bottlenecks (plumbers, electricians, energy) and the difficulty of measuring productivity in a complex, evolving economy. The most actionable takeaway is that the "price" of AI will likely shift from flat subscriptions to usage-based models, necessitating a new corporate role: the "Chief Token Officer," who manages the trade-off between human labor and AI compute costs.