The S&P 500 is only a stone's throw from its all-time high, but beneath this calm surface the market's internals have become severely fractured. What is propping up the index now is essentially a single pillar 鈥?the "Magnificent Seven" tech giants 鈥?and that is precisely where the market's biggest risk lies.
Jonathan Krinsky of technical analysis firm BTIG warned in his latest report titled "Last Man Standing" that the Magnificent Seven are the only group within the S&P 500 that has yet to show obvious signs of fatigue, and also the most crowded trade in the current market. He noted that once this trend begins to reverse, it will very likely become the trigger for a sustained decline in the S&P 500.
Meanwhile, data from Michael Cembalest of JPMorgan Asset Management shows that 60% of S&P 500 constituents have fallen more than 20% from their respective all-time highs, entering technical bear markets, while the index itself still hovers near record levels.
This divergence between the surface and the internals has made the market's fragility increasingly pronounced. Analysts point out that when defensive trades and growth trades are concentrated in the same seven stocks, the market has essentially lost room for rotation; and the persistent decline in the semiconductor sector is unfolding along the path of the 2000 tech bubble burst, suggesting that the real pressure may be released in a concentrated fashion in late October.
The Magnificent Seven Can't Hold It Alone, Crowded Positioning Plants Hidden Dangers
Krinsky's core judgment is simple and direct: the Magnificent Seven are currently the only major sector still making new highs, but they are also the most concentrated in terms of positioning.
Looking at market data, the Roundhill Magnificent Seven ETF (MAGS) has broken above both its 2025 and May highs to set a record, while surrounding sectors have broadly pulled back. On Tuesday, even as semiconductor stocks fell again, the Magnificent Seven as a whole rose about 1%, single-handedly propping up the S&P 500.
Yet this "sense of safety" is precisely the problem. Hedge funds have treated the Magnificent Seven as a safe haven from Treasury yields above 5% and geopolitical risk, flooding in and driving concentration to historic extremes. Krinsky pointedly noted: when defensive trades and momentum trades point to the same seven stocks, the market has nowhere left to rotate. A "safe-haven asset" that everyone holds is, by nature, no longer a safe haven it is a beautifully packaged crowded trade.
Semiconductor Trend Replicates 2000, Accelerated Decline May Come at Month-End
Another warning signal from BTIG comes from the semiconductor sector. Since the summer, Krinsky has been tracking the comparison between the Philadelphia Semiconductor Index (SOX) and its 1999鈥?000 trajectory, calling the resemblance "uncomfortably" close.
Under this analogy, the semiconductor sector is currently near a short-term low and is about to see a brief rebound 鈥?the CPI data release may serve as a catalyst 鈥?followed by an accelerated decline in late October. And late October happens to be the window when major hyperscalers report earnings in concentrated fashion.
Krinsky also noticed another signal: the relative strength ratio between software and semiconductors (IGV/SMH) has reclaimed its 200-day moving average for the first time in 16 months. He believes that if this ratio further breaks through the 0.20 threshold near its summer high, it would "confirm a major inflection point," meaning market leadership would shift from semiconductors to software 鈥?and software is not where the market is heavily positioned right now.
42 AI Stocks Drive Sixty Percent of Gains, Everything Else Is Just Background
Michael Cembalest of JPMorgan Asset Management, in his latest report "Califoreboding," uses the title "One Trick Pony" to lay out with data just how extreme market concentration has become.
Since January 2024, 42 AI-related stocks have contributed roughly 60% to 80% of the S&P 500's earnings growth, price returns and capital expenditure growth. Specifically, these 42 stocks contributed 67% of the index's price return, 58% of earnings growth, and as much as 81% of capital expenditure and R&D spending growth.
At the same time, although the S&P 500 is near its all-time high, 60% of its constituents have fallen more than 20% from their respective records. Cembalest's tally shows that since 1990, it has been extremely rare for such a large proportion of individual stocks to suffer deep drawdowns while the index sits less than 1% from its record, with the September 25 reading sitting at the extreme left tail of the historical distribution. Moreover, only 30% of S&P 500 constituents have outperformed the index over the past three years, among the lowest levels since 1991, with the only comparable trough occurring in 1999鈥?000.
AI Bonanza Flows to the Shovel Sellers, While the Diggers Have Yet to Strike Gold
Cembalest's data also reveals another side of the AI boom: profits are highly concentrated among the suppliers of AI infrastructure rather than its users.
Since March 2023, the trailing earnings of 55 "AI supply chain" companies have roughly quadrupled, while the earnings of 190 "AI-heavy adoption" companies grew only about 25%, and the earnings of companies with the lowest AI adoption were essentially flat or slightly lower.
This pattern closely resembles the internet bubble era: those selling the shovels made a fortune, while the diggers have yet to find the gold. The concentration effect has also spread to the real economy 鈥?tech's contribution to U.S. GDP growth climbed from nearly zero in 2022 to about one-third, while only about 20% of U.S. states are growing faster than the national average, close to the lowest level since 1979.
In the construction investment space, spending on data centers and power infrastructure has risen 38% since January 2024, while manufacturing construction fell 28%, commercial real estate dropped 13%, and residential construction declined 6%. More notably, pre-construction planned data center capacity has fallen back to about 49 gigawatts from a peak of roughly 62 gigawatts, signaling that this sole engine of capital expenditure growth has begun to scale back its plans.
Bulls Still Have Support, but the Earnings Foundation Is Narrowing
Cembalest has not taken an explicitly bearish stance. He notes that since January 2024, the rally in global tech stocks has been driven mainly by earnings growth rather than valuation expansion, and "a market driven by earnings growth, in my experience, is not prone to rapid declines." In addition, despite rising short-term rates, corporate net interest expense as a share of profits remains at historically low levels.
However, the sources of these earnings are highly concentrated 鈥?mainly the profits of AI supply chain companies, and these profits come in large part from the capital expenditure of hyperscalers, whose expansion increasingly relies on debt financing. The earnings themselves are real, but the question is whether these customers' spending can be sustained. According to Goldman Sachs estimates, the third quarter is likely to be the dual peak for both S&P 500 earnings growth and hyperscaler capital expenditure growth.
Combining the analyses from BTIG and JPMorgan, the current market picture is clear and stark: 42 stocks contributed two-thirds of returns and four-fifths of capital expenditure growth; 60% of constituents have fallen into bear markets; only 30% of stocks have outperformed the benchmark; defensive trades and growth trades are concentrated in the same seven names, with hedge fund positioning at record highs; and the semiconductor sector has already declined along the 2000 path, with the acceleration phase potentially arriving in late October around the hyperscaler earnings season.
Krinsky's conclusion remains restrained: "Fighting new highs is never easy." The Magnificent Seven's charts still point upward. But as he warned: when the market is down to its last leg, it doesn't take much force to push it over.
Comments