The correction already happened. The pivot is only a name.

Evercore counted 115 stocks in the S&P 500 that spent the past six months trading against their own index: falling when it rises, rising when it falls. Beta measures how a stock moves relative to the market, and for these 115 the number has gone negative ^[1]. The index they are moving against just entered the fifth year of its bull market ^[2]. The ledger behind those two facts fills in fast. An equal-weight version of the S&P 500, where every company counts the same instead of the biggest carrying the most, is in its longest losing streak against the cap-weighted index since 2004 ^[3]. Market breadth, the share of stocks actually rising with the market, is the lowest since the dotcom bubble, by Goldman Private Wealth's count ^[3]. Roughly 40 to 45 percent of the index now trades opposite it over three months, and AllianceBernstein found an unprecedented share of negative-beta U.S. stocks in July ^[1]. The biggest AI-linked companies carry 70 to 80 percent of the index's earnings growth while nearly half the companies in it are down on the year ^[2]. If the correction already happened, quietly, inside the index, away from the headline number, then what is the pivot the warnings keep announcing? Fund managers have started selling small-cap value under a new name.

small cap value is the anti-AI trade — Simon Evan-cook

The pitch comes with a built-in caveat: the rotation only works if it isn't set off by a broader crisis or war ^[4]. The rotation itself has been declared three times since spring and stalled three times. In April investors moved money out of AI, knocking tech-software funds down more than 20 percent while staples and industrials gained ^[5]. In July small caps surged past the S&P 500 and the Nasdaq ^[6]. By mid-October the small and micro caps were back at oversold levels, while mega-cap AI names still propped up the cap-weighted index ^[7]. The professionals who actually sold AI in the second quarter didn't buy the advertised small caps. Viking Global sold out of Apple, Appaloosa exited Microsoft, Two Sigma liquidated Amazon and Tesla, Leon Cooperman sold his Alphabet, and Third Point exited Meta. The money went to Costco, Walmart, Home Depot, and Eli Lilly instead ^[8]. Even the rout got bought rather than traded around. On the worst day of the eight-session semiconductor selloff, TPO portfolio manager Uttam Dey was redeploying cash into AI infrastructure stocks, arguing the bond market showed equities were near a bottom ^[9]. September had already shown the same reflex, with software stocks recovering on strong earnings once the pause calls made the catastrophic scenarios look unlikely ^[10]. Both camps have quietly stopped arguing about the next few quarters. Goldman projects a poor 10-year outlook for equity returns from high multiples plus concentration ^[11]. Vanguard projects value stocks as the best-performing asset class over a decade, at 6.4 to 8.4 percent annualized, against 3.6 to 5.6 percent for large-cap growth; Fidelity expects small caps to beat large caps over five to ten years, citing a Shiller price-to-earnings ratio, a smoothed long-run valuation gauge, near its all-time high of 41.59 ^[12]. On the spending side there is no pivot at all. Hyperscaler capital spending is still rising into 2026 and 2027, and it is increasingly funded with debt as the cash-to-capex ratio, cash on hand relative to what is being built, falls ^[11]. The five biggest AI firms are issuing debt to fund that buildout ^[13]. By late July the 30-year Treasury yield had reached roughly 5.28 percent, its highest since 2007, while the Federal Reserve held its own rate steady ^[13]. Analyst Casey Sprake describes what happens next as a closed loop.

the bond yield curve is actually doing the the work for the Fed in that — Casey Sprake

The loop has a price tag. Bank of America calculates the weighted average cost of capital, the return a project must clear to be worth funding, for the AI hyperscalers has jumped to 10 percent, from 4 percent in 2023, raising the bar every future AI project has to clear ^[14]. None of this is welded together yet. The analyst warning the debt could cascade concedes the credit trouble so far is isolated to specific names, not systemic ^[15].

my fear is all of a sudden that trickle of selling becomes a bucket getting dumped, and the market doesn't have that liquidity, and that's when it drags everything down. — Issa Tchiroma

TSMC held its revenue outlook even as the regional selloff raged around it ^[16]. And the analysts pointing to dotcom-era dispersion concede the difference: today's leaders are mature businesses with established revenue, not the startups of 1999 ^[1]. The tell is in the timing. The anti-AI trade got its name the same week small caps slipped back to oversold. The money that genuinely left AI announced nothing and bought groceries. A trade that is actually happening does not need a name; the name marks where the money isn't.