Wall Street keeps hitting new highs, but a growing number of institutional voices are beginning to list artificial intelligence (AI) itself as the biggest threat facing global markets. The S&P 500 index is at the center of this debate, and its concentration explains why.
Tail risk is an event with low probability but severe consequences. Bank of America's July global fund manager survey showed that 45% of respondents viewed the AI bubble as the biggest tail risk, up from 28% the previous month. This proportion surpassed the second wave of inflation, becoming fund managers' primary concern. Meanwhile, long positions in global semiconductors were seen as the most crowded trade. Respondents also noted that hyperscalers' spending on AI infrastructure was most likely to trigger a credit event.
Analyst Mac10 issued a warning on August 8, believing that forward earnings growth is accelerating at an unprecedented pace, simply because companies are pouring unprecedented cash into AI. Such spending often shows up as a one-time boost in the income statement, rather than sustainable operating performance. The Bank for International Settlements also warned that the spending spree of tech giants could evolve into a prolonged investment slump. The top five hyperscalers are expected to invest more than $1 trillion in 2025-2026.
Household equity exposure is higher than in previous cycles, so any sharp decline would have a greater impact than the bursting of the internet bubble.
JPMorgan Global Research estimates that the top 20 stocks account for about 50.8% of market capitalization, a concentration with no precedent in modern history. In half a century, the index has never been so dependent on a few companies. Buying the market increasingly means buying the AI trade, regardless of how the remaining 480 companies perform.
Capital commitments are still expanding. Goldman Sachs estimates that by the end of 2026, annualized AI-related spending could exceed $800 billion. Morgan Stanley expects AI infrastructure investment to approach $3 trillion by 2028, with more than 80% of that yet to occur.
A stress test was conducted this summer. According to TradingView data, the Nasdaq fell nearly 10% from its June peak to the end of July, then rebounded nearly 9% in early August, hitting a record high. Momentum stocks proved fragile; SanDisk and Western Digital rose about 396% and 145% respectively this year, but both saw "sell the news" pressure during earnings season.
However, the bullish view is that earnings are sufficient to support valuations. Goldman Sachs found that 64% of S&P 500 companies beat consensus expectations by at least one standard deviation. BlackRock explicitly denied the bubble theory, arguing that current leaders have real earnings, strong balance sheets, and most investment comes from free cash flow.
If the market needs a case study of AI concentration risk, July provided one. The situational awareness hedge fund founded by former OpenAI researcher Leopold Aschenbrenner once managed $45 billion in assets, but due to heavy losses in AI infrastructure stocks such as SK Hynix, it was forced to sell its entire public portfolio to Ken Griffin's Citadel. Six days later, Citadel took over shares once valued at $16 billion, one of the largest emergency equity deals in Wall Street history. A cascade of margin calls shrank the fund's assets from $45 billion to about $10 billion within weeks. However, after the near-collapse, Aschenbrenner invested another $400 million in a private company, bringing his cumulative investment in the unnamed target to $500 million, while retaining private equity stakes.
This event does not prove that the AI trade is over, but it reveals how concentration, leverage, and low liquidity can destroy a portfolio before the long-term thesis is realized—a fragility that is now embedded in the S&P 500 itself at an index scale.
