Federal Reserve Says U.S. Stock Valuations Near Internet Bubble
The Federal Reserve stated in its recent monetary policy meeting minutes that valuations in some asset classes "seem elevated," and the market's pricing of risks associated with overvalued tech assets remains relatively loose.
The S&P 500 Shiller price-to-earnings ratio is approximately 42.5 times, close to the 44.2 times peak during the 1999 internet bubble; this metric measures current stock prices against inflation-adjusted earnings over the past decade, reflecting that market prices have significantly outpaced long-term earnings bases.
The rise in U.S. stocks is concentrated among large tech companies related to artificial intelligence, with investors continuing to assign higher valuations to long-duration growth assets based on expectations of future productivity improvements, increased corporate software spending, and expanded demand for computing power; however, this also makes prices more sensitive to downward adjustments in earnings expectations.
The Federal Reserve did not directly define high valuations as systemic risk in the meeting minutes but used the phrase "some asset classes are overvalued"; this wording indicates that policymakers are monitoring asset prices in relation to financial stability risks.
The Nasdaq Composite Index closed at 26,289.71 points on August 18, remaining in a high range. High valuations do not automatically mean an immediate market reversal, but when risk-free interest rates, earnings growth rates, or AI capital expenditure expectations reverse, valuation compression can amplify stock price volatility.
Market mechanisms show that funds are primarily flowing into high-duration assets such as large-cap tech, semiconductors, cloud computing, and AI infrastructure, with buyers mainly consisting of passive index funds, growth-oriented funds, and allocations chasing AI themes; if earnings realization lags behind valuation expansion, funds will first withdraw from high-valuation, low free cash flow assets, benefiting defensive sectors, short-duration assets, and high-rated bonds.
Source: Public Information
ABAB AI Insight
During the internet bubble from 1999 to 2000, capital similarly first reached a consensus that "the internet would reshape business," then front-loaded valuations far exceeding current earnings levels. Subsequently, the Nasdaq peaked in March 2000, and many cash-strapped internet companies exited the market; however, companies like Amazon and Cisco, which had infrastructure or scale advantages, became long-term survivors. The key difference in the current AI market is not the narrative, but that the leading entities are mostly large platform companies with substantial cash flow, cloud businesses, and capital expenditure capabilities.
The capital path has shifted from venture capital-driven early internet financing to a heavy asset competition dominated by super platforms. Large cloud service providers secure supply through data centers, GPU procurement, proprietary models, enterprise distribution channels, and long-term power contracts; rising valuations also lower financing costs for leading firms, creating a positive feedback loop of "capital expenditure—revenue expectations—valuation—reinvestment."
Historical comparisons are more akin to the cycles of railroads, telecom fiber optics, and internet backbone networks: infrastructure is typically built in concentration before demand is fully realized, followed by price wars and idle capacity that eliminate marginal participants. Around 2000, U.S. telecom companies over-invested in fiber networks, leading to the collapse of companies like WorldCom under demand and financial pressure; today, the AI industry is still transitioning from infrastructure expansion to commercial validation, where model capabilities do not equate to sustainable, distributable profits.
This essentially represents a simultaneous transfer of pricing power and capital concentration: AI training costs, computing power supply, data center construction, and enterprise distribution capabilities will raise competitive barriers, and capital will preferentially concentrate in companies that can simultaneously secure chip access, cloud platforms, customer channels, and cash flow. Valuation risk depends not only on whether the technology is real but also on whether future profits will be captured by infrastructure suppliers, platform operators, or end-application companies.
ABAB News · Laws of Cognition
Technology changes the world, cash flow determines winners
High valuations are not a risk; unmet expectations are
Infrastructure builds prosperity first, pricing power captures profits later