The S&P 500 Index keeps setting fresh record highs, yet internal market divergence, rising financing costs and the risk of slowing AI capital expenditure growth are quietly building up.

Goldman Sachs veteran macro trader Bobby Molavi warned that the current US equity market has become increasingly dependent on the strength of a handful of large technology stocks, while professional investors find it hard to profit from the index's advance, meaning market fragility is on the rise.

Index hits new highs, but the market is increasingly fragile

Molavi described the current US equity market as "running desperately yet standing still." The S&P 500 keeps hitting new highs, yet net exposure in US fundamental long-short strategies has already dropped to about 46%, close to the low seen during the April 2025 market selloff, reflecting how difficult it is for hedge funds to profit from this rally.

Market divergence is a key reason. The report shows that the gap between the three-month realized volatility of individual stocks and the realized volatility of the index has widened to 26 to 28 percentage points, exceeding the historical high of around 20 percentage points set in 2009. Sharp swings in individual stocks offset one another, keeping the index superficially calm while making trading harder.

Current market concentration has surpassed the peak of the 2000 dot-com bubble. The report shows that 10 core AI-related stocks now account for roughly 40% of the S&P 500's total market capitalization. The best-performing AI-related stocks have risen an average of 58% year to date, far outpacing the S&P 500's 13% gain and the Nasdaq's 22% gain, while the median S&P 500 constituent has actually fallen 15%. In addition, just five stocks have a combined market value equal to that of another 434 stocks, highlighting how heavily the US equity rally relies on a small number of large companies.

It is worth noting that liquidity has deteriorated further. Average daily order book liquidity in the US equity market has fallen 21% since 2023, from about US$14 million to US$11 million; order book depth per US$1 billion of trading volume has dropped from US$162,000 in 2017 to US$11,000, a decline of about 93%. Over the same period, average daily volume rose from about 6.5 billion shares to nearly 19 billion shares. Rising volume alongside shrinking order book depth means the market's capacity to absorb large trades has weakened, making individual stocks more prone to sharp swings on sudden news.

AI capex nears an inflection point as stock prices, earnings and financing risks shift at the same time

Molavi estimates that the entire AI ecosystem — spanning large technology companies, cloud service providers, AI labs, chips and computing infrastructure — now represents US$25 trillion to US$30 trillion in related assets, whose value depends heavily on continued infrastructure investment.

Historical experience shows that stock prices typically peak before capital expenditure does. Among the three key signals Molavi identifies, divergence between stock prices and actual earnings and a credit market reassessment of AI borrowers' risk have already appeared, while a slowdown in capex growth has yet to be confirmed.

First, forward price-to-earnings ratios may understate true valuations. Report data shows that in March 2000, the S&P 500's forward P/E was 25.2 times, but based on actual realized earnings it reached 39.3 times, and on a GAAP basis it was about 60 times. In 2007, the forward P/E was only 15.1 times, yet based on realized earnings it reached 31.6 times, and on a GAAP basis it exceeded 100 times. Molavi cautioned on this basis that whether forward valuations are reasonable ultimately depends on whether corporate earnings can be delivered.

Second, AI financing risk is being repriced. The report notes that more than US$500 billion of debt issuance has been used this year to support AI infrastructure buildout. Yields on several well-known AI financings have continued to rise, and investment-grade and high-yield credit spreads have also begun to widen, indicating that creditors are raising the price they charge for related financing risk.

Third, capex growth may be about to peak. Goldman Sachs forecasts that hyperscaler capex growth will peak at 116% year over year in the third quarter of 2026, then fall to just over 20% by the end of 2027. A peak in the growth rate does not mean total capex will immediately decline, but if the market had already priced in continued acceleration in investment, a slowdown in growth could still change investors' expectations for future returns.

Changes in the financing environment are adding further pressure. The number of monthly US equity offerings fell from 65 in May to 24 in September, while quarterly issuance totals dropped from US$252 billion in the second quarter to US$93 billion in the third quarter. Follow-on offerings have performed weakly, with some deals forced to be pulled. Although Broadcom and SpaceX are still preparing new rounds of financing, high interest rates and rising uncertainty over investment returns are making buyers more cautious than in the first half of the year.

Buybacks still support US equities as oil and bond market risks build in tandem

As financing tightens, share buybacks continue to provide support for US equities. S&P 500 companies' buybacks rose 20% year over year in the second quarter, and full-year US buyback authorizations are on track to set a record high. Still, it remains to be seen whether buybacks can offset market doubts about earnings delivery and AI investment returns.

At the macro level, pressure in the French bond market and energy market is also building. Molavi noted that the spread between French and German 10-year government bonds has reached 140 basis points, the highest level since 1990, reflecting market concerns about France's fiscal and political risks and warning that pressure could spread to other eurozone countries.

In energy, repeated geopolitical tensions, high shipping and insurance costs, and refining capacity constraints are eroding the market's buffer. The report shows that jet fuel prices range from US$1,100 to US$1,400 per tonne depending on contract terms, well above the long-term average of about US$700. If fuel costs push airfares even higher, it could weigh further on consumer budgets and prompt some people to choose closer, cheaper trips.

Molavi's warning is not an assertion that AI investment has already peaked, but rather a signal that the market is approaching a critical phase in which investment returns will be re-examined: stock prices are diverging from actual earnings, the credit market is starting to charge more for AI borrowers' risk, and capex growth may also be turning toward a slowdown.