Table of Contents
ToggleBy 2026, the Shiller price-to-earnings ratio (CAPE) , a classic valuation metric for the US stock market, has climbed to between 39.5 and 41.7 times , reaching its highest level in the past 25 years. This figure is only lower than the extreme value of about 44 times at the peak of the internet bubble in 1999.
This means that, measured on a historical scale, the S&P 500 index has now entered a "highly inflated" range.
AI-driven repricing
The core driver of this round of valuation surge is not the frenzy of 1999 fueled by the ".com" suffix and the dream of wireless telephony, but rather a comprehensive repricing triggered by the AI (Artificial Intelligence) concept . Tech giants related to semiconductors, cloud computing, and AI infrastructure, led by NVIDIA, Microsoft, Amazon, and Meta, have continued to rise, with a few large tech stocks accounting for a significant portion of the S&P 500's gains.
Market observers worry that the current market trend shares structural similarities with the 1999 dot-com bubble, featuring a dominant technology narrative, inflated market capitalization of a few star companies, and a widespread belief that "this time is different ." However, simply categorizing the current market as a "bubble" may be an oversimplification, as the fundamentals supporting these high valuations are vastly different from those of the past.
AI Giants vs. 1999
Unlike many tech companies in the past that were not yet profitable and relied solely on their "user growth curve" to convince investors, today's leading AI companies generally possess the following characteristics:
- Strong free cash flow : NVIDIA's annual free cash flow has exceeded $80 billion, and Microsoft's cloud business (Azure) is approaching $300 billion in annual revenue.
- Mature Business Model : Unlike the "expand first, figure later" approach of 1999, AI giants now derive their revenue primarily from subscription-based SaaS, cloud-based usage-based billing, and enterprise-level contracts.
- High profit margins : Magnificent Seven's consolidated operating profit margins generally fall between 30% and 35% , far exceeding the approximately 17% of the Nasdaq 100 in 1999.
Furthermore, the current AI investment boom has expanded from the chip level to suites, data centers, energy (ultra-high voltage power transmission, micro nuclear reactors), power grids, and enterprise-level AI applications. This presents bubble theorists with an awkward question: if this is a bubble, it may be the "most fundamentally sound" bubble in history.
Concentration risk and return compression
Despite stronger fundamentals than in 1999, analysts still point to three major risk factors:
Concentration risk : Excluding the Magnificent Seven, the remaining 493 stocks in the S&P 500 underperformed significantly. This "seven giants dominating the market" structure makes the market particularly sensitive to pullbacks in individual giants.
Interest Rate and Commercialization Uncertainties : The Federal Reserve's interest rate environment (Fed Funds Rate approximately 4%–4.25% ) is not yet fully priced in. If interest rates rise above 4.5% in the future due to a rebound in inflation, the return on capital expenditures (CapEx) for AI infrastructure may face a revaluation. Simultaneously, if the commercialization of AI falls short of expectations—for example, if corporate adoption slows or ROI is delayed—pressure for valuation compression will quickly emerge.
Historical Return Compression : Historical data shows that after the Shiller price-to-earnings ratio enters a historically high range (CAPE > 30), the real return of US stocks over the next ten years (after adjusting for inflation) typically tends to be in the low single digits (approximately 1%–3%) . This means that without a "hard landing," the high valuation itself will compress future annualized returns.
How should we view this?
The Shiller P/E ratio is not a panacea, because while the US stock market did experience a historic compression of returns after 1999, it also took nearly a decade for the CAPE ratio to recover from its decline from 44 times in 2000 to around 13 times in 2009. The current market may not follow the exact same trajectory, especially when the productivity benefits brought by AI are real and quantifiable.
But history has taught us that high valuations do not necessarily mean a bubble; they simply mean that future returns will be more demanding. In 2026, with CAPE approaching 40, investors should perhaps not ask "Will this time be different?" but rather "Will this time be different enough to justify a 40x valuation?"
Related reports
