Fed Vice Chairman: Analyzing the AI Bubble from Four Dimensions

Author: Zhang Feng

Artificial intelligence (AI) is reshaping the global economy and financial landscape at an unprecedented pace. As the capital markets' enthusiasm for AI-related companies continues to surge, an inevitable question arises: Are we witnessing a speculative frenzy similar to the internet bubble of the late 1990s?

In 2025, Federal Reserve Vice Chairman Philip N. Jefferson systematically elaborated on his comparative analysis of the current AI boom and the internet bubble era at the Financial Stability Conference of the Cleveland Federal Reserve Bank, proposing four key indicators for determining whether there is a bubble in AI. This speech not only reflects the cautious observation of emerging technologies by the world's most important central banks but also provides market participants with a clear framework for rationally assessing the AI boom.

I. Observations of the Federal Reserve: Dual Mandate and Financial Stability

All policies and observations of the Federal Reserve revolve around its statutory “dual mandate”—maximizing employment and price stability. Jefferson clearly stated that the impact of artificial intelligence must be assessed from this fundamental task. This means that the Federal Reserve's concern with AI is not only about its technological breakthroughs or market performance, but also about how it affects overall employment levels, labor productivity, economic growth potential, and inflation trends.

From an employment perspective, AI exhibits a dual effect. On one hand, it promotes employment by enhancing work efficiency and creating new positions (such as AI research and development, deployment, and maintenance); on the other hand, its automation substitution effect may lead to the shrinkage of certain professions, particularly impacting younger and less experienced workers more significantly. Jefferson pointed out that if AI merely replaces the existing workforce without simultaneously creating new positions, it could trigger a short-term economic slowdown. This dynamic balance of “substitution and supplementation” is key to assessing the structural impact of AI on the labor market.

From the perspective of price stability, AI improving productivity helps reduce production costs, putting downward pressure on prices. Applications such as efficient resource allocation, supply chain optimization, and decision support may suppress inflation. However, at the same time, the construction of AI infrastructure (such as data centers) raises the prices of inputs like land and energy, and rising salaries for AI talent may also lead to cost-push inflation. This dual effect makes the net impact of AI on inflation highly uncertain, requiring continuous monitoring.

To achieve dual missions, a robust and resilient financial system is crucial. The Federal Reserve continuously monitors systemic risks through its semiannual Financial Stability Report (FSR). The latest survey shows that 30% of market contacts view “the shift in attitudes towards AI” as a significant risk to the financial system, a substantial increase from 9% in the spring. This seems to warn that if market optimism about AI suddenly reverses, it could trigger tightening financial conditions and an economic downturn. Therefore, the Federal Reserve's inclusion of AI in its financial stability monitoring framework is intended to guard against asset bubbles and financial vulnerabilities that may arise from technology booms.

2. Monitoring Framework: FSR and Market Sentiment Tracking

The Federal Reserve's monitoring of AI is not conducted in isolation but is embedded within its overall financial stability assessment framework. The FSR not only focuses on traditional risks such as leverage, asset valuation, and financing risks but also incorporates structural changes brought about by emerging technologies. Jefferson emphasized that policymakers must distinguish between “cyclical fluctuations” and “structural changes,” and AI is likely to belong to the latter. This means that the productivity gains brought by AI may alter the relationship between employment and inflation, thereby affecting the transmission mechanism of monetary policy.

Market sentiment is one of the key focuses of FSR. Surveys show that nearly one-third of market participants have become aware of the potential risks of an AI sentiment reversal. This consensus itself could become a “self-fulfilling prophecy” — once the optimistic narrative shifts, rapid capital withdrawal could lead to severe asset price adjustments. Compared to the internet bubble period, the speed of information dissemination today and the prevalence of algorithmic trading may amplify market volatility. Therefore, the Federal Reserve's tracking of sentiment indicators is essentially an early warning of potential systemic risks.

In addition, the application of AI in the financial industry itself also brings new monitoring challenges. High-frequency trading, intelligent investment advising, risk models, and other AI tools improve efficiency while potentially introducing new homogenization risks and procyclicality. The Federal Reserve is strengthening the identification and assessment of these emerging risks by expanding its analytical toolkit (including the use of AI technology itself).

Three to Four Core Indicators: The Touchstone for Judging the AI Bubble

Jefferson distilled four key differences by comparing the current AI craze with the internet bubble of the late 1990s, which can serve as core indicators to judge whether there is a serious bubble in the current AI field.

(1) Profit Basis: From “Story-Driven” to “Profit Support”

During the internet bubble period, many companies went public solely on the concept of “.com”, lacking sustainable profit models, with meager or even zero revenues, relying on external financing and market enthusiasm to maintain operations. In contrast, leading companies in the current AI field (such as some tech giants) generally have solid and diverse profit channels. They not only generate revenue directly through AI services but also deeply integrate AI into existing product systems, enhancing the competitiveness of their core business. This development model of “profit support” makes AI investments more fundamentally sound, reducing the space for pure speculative trading.

However, Jefferson also pointed out that the activity in the private equity market may partially obscure the profitability challenges faced by early AI companies. A significant amount of venture capital is flowing into AI startups, which, although not publicly listed, have high valuations. If they are unable to achieve profitability in the future, they may still pose a risk. Therefore, the observation of profitability indicators must take into account both public and private markets.

(2) Valuation Level: Price-to-Earnings Ratio Relatively Moderate

During the peak of the internet bubble, the price-to-earnings ratios of internet companies often reached hundreds or even thousands of times, reflecting the market's irrational optimism about long-term growth. Currently, although the stock prices of AI concept companies have surged significantly, their price-to-earnings ratios are still far below historical peaks. This to some extent indicates that while investors are chasing AI, they are still anchoring to the actual profits and cash flows of companies.

Of course, the reasonableness of valuation needs to be comprehensively judged in conjunction with industry characteristics and growth stages. As a general-purpose technology, AI has immense potential for long-term value creation, and a moderate premium is reasonable. However, if valuations rise too quickly away from fundamentals, it may still breed bubbles. The Federal Reserve's attention to valuation indicators is precisely to discern the rational components of market enthusiasm from overheating signals.

(3) Number of listed companies: Limited speculative breadth

From 1999 to 2000, over 1,000 internet companies went public, creating a speculative frenzy characterized by “flowers blooming everywhere,” where even a name change to include “.com” could drive up stock prices. Currently, there are about 50 publicly listed companies explicitly classified as “core AI enterprises” (based on specific criteria), which is far fewer than during the internet bubble period. This indicates that market speculation is relatively concentrated and has not yet spread throughout the entire market.

But Jefferson also reminded that the private equity market may hide a large number of AI startups that, while not publicly traded, are actively involved in financing activities. If these companies go public in large batches in the future or if the financing environment changes dramatically, they could become new sources of instability. Therefore, the “number of companies” indicator needs to be observed dynamically, covering both public and private sectors.

(4) Financial Leverage: Low debt dependence

During the internet bubble period, many companies relied on equity financing, with limited debt leverage, which somewhat reduced the direct impact of the bubble burst on the financial system. Currently, AI companies similarly rely less on debt financing, which helps to limit risk transmission. However, recent trends indicate that to support massive investments in AI infrastructure (such as data centers and computing clusters), some enterprises have begun to increase bond issuance and credit financing.

Jefferson specifically pointed out that as AI expands from software to hardware infrastructure, the demand for capital investment has surged sharply, which could lead to a gradual increase in leverage. If AI sentiment reverses, highly leveraged companies will face greater debt repayment pressure, thereby spreading risk to broader economic sectors through credit channels. Therefore, leverage indicators need to be closely monitored for their evolving trends.

4. Insights for Market Practitioners

Jefferson's discourse not only provides an analytical framework for policymakers but also brings important insights to investors, businesses, and researchers:

First, observing issues should start from the fundamental tasks of the observer. Investors should go beyond short-term market sentiments and deeply analyze the substantial impact of AI technology on the fundamental aspects of enterprises (profitability, cost structure, competitive barriers). Enterprises should focus on how AI enhances their productivity and long-term competitiveness, rather than blindly chasing concepts.

Second, distinguish between cyclical fluctuations and structural changes. AI represents a technological revolution that could last for decades, and its impact is structural. In market fluctuations, it is important to differentiate between long-term trends and short-term noise, to avoid misjudging structural opportunities as cyclical bubbles, or vice versa.

Third, pay attention to the overall market response and systemic risks. The rise of a single company or sector does not necessarily constitute a bubble; it is necessary to assess the overall market valuation level, capital concentration, leverage situation, and emotional consistency. Particularly, be wary of signs that the AI narrative is shifting from “profit support” to “story-driven.”

Fourth, make good use of analytical tools, including AI itself. AI technology can be used to more accurately assess market risks, corporate value, and economic impact. Practitioners should actively utilize data analysis, machine learning, and other tools to enhance decision-making quality, while also being aware of the new risks that may arise from model homogeneity.

5. Continuously, multidimensionally, and dynamically participate with reason and passion

Jefferson's final conclusion is relatively cautiously optimistic: based on comparisons across four dimensions—profitability, valuation levels, number of companies, and financial leverage—the current AI boom is significantly different from the internet bubble, and the likelihood of a severe collapse akin to that of the late 1990s is low. The development of AI is rooted in a number of mature companies with solid profitability, and the overall financial system is quite resilient.

However, uncertainty still exists. The long-term impact of AI on employment, inflation, and productivity still needs time to be validated; market sentiment may reverse; the activity level in the private equity market may mask risks; the potential for infrastructure investment to increase leverage is worth being vigilant about. Therefore, the Federal Reserve will continue to monitor the development of AI, ensuring that it unfolds in a stable and resilient financial environment, ultimately serving the fundamental goals of maximizing employment and price stability.

For the market, Jefferson's analysis provides a toolbox for rational assessment of AI investments. In the wave of technological revolution and capital enthusiasm, maintaining clarity, distinguishing essence from appearance, and focusing on long-term value may be the best posture to avoid bubbles and embrace change. Is AI a bubble? The answer is not a simple yes or no, but lies in continuous, multidimensional, and dynamic observation and judgment.