The Magnificent 7 is no longer beating the market. A fresh report from Apollo Global Management chief economist Torsten Sløk reveals that with May 1, 2026 set as the base, the Mag 7 index dropped to roughly 91 by the end of June, while the S&P 500 stood at about 102. The rest of the index—S&P 493—rose to around 107, meaning the seven tech giants actually underperformed the average stock.
Market-Cap Share Shrinks from 33% to 28%
The Mag 7's combined weight in the S&P 500 fell from about 33% at end-2025 to 28%. Though five percentage points sounds modest, the absolute scale is staggering: the group still holds roughly $20.6 trillion in market cap, exceeding the total stock markets of Japan, the UK, and Canada combined (about $17 trillion). Nvidia alone, at $4.7 trillion, is larger than the entire equity market of Canada, the UK, or France. This shift signals that the market's obsession with concentrated bets is easing.
Performance divergence inside the group is stark. Measured from October 1, 2025, Alphabet surged as much as 60-65% and still holds a ~40% gain; Amazon, Nvidia, and Apple show modest positive returns. Tesla, Microsoft, and Meta, however, have flipped negative, with losses between 20% and 30%. Treating the seven as a uniform AI basket is proving costly.
Free Cash Flow Bleeds: AI Data Centers Gobble Up 75% of Operating Cash
The next-12-month free cash flow of hyperscalers (Amazon, Microsoft, Google, Meta, Oracle) has collapsed from a peak of about $290-300 billion in 2024 to just $50-65 billion by mid-2026, with some names turning negative. Free cash flow—the cash left after necessary investments for maintenance and expansion—is the dry powder for dividends, buybacks, or further reinvestment. That dry powder has been cut by more than 80%.
The culprit: AI capital expenditure. Capex covers building data centers, buying Nvidia GPUs, and installing power and fiber. The ratio of capex to operating cash flow for these firms jumped from about 40% in 2024 to roughly 75% by March 2026. For every $100 earned, $75 goes into data centers. Annualized spending on data center construction has reached around $50 billion, surpassing office building construction ($44 billion and falling). Investment growth remains brisk at 28-35% annually, but whether this infrastructure build will yield commensurate commercial returns in a foreseeable time frame remains an open question.
Valuation Premium Hits Decade Low: AI Certainty Being Repriced
The Mag 7's forward P/E stands at about 24-25x, versus S&P 493's ~21x. The premium—historically 50-75% over the past decade—has collapsed to just 12%. The market is no longer willing to pay a fat premium for the Mag 7 halo. Earnings growth convergence tells a similar story: in 2024 Mag 7 earnings grew 37% (S&P 493: 5%); in 2025: 23% vs 11%; 2026 estimate: 36% vs 20%; by 2027 the gap shrinks further to 19% vs 15%. Margins remain wide (23-24% vs 10%), but high margins combined with surging capex erode the efficiency of converting profits into shareholder returns.
Apollo's report deliberately avoids the word "bubble," stating only that "markets are rotating toward quality and free cash flow." That deliberate understatement leaves room for two paths: if AI capex eventually delivers, this is a healthy rebalancing; if returns disappoint, the quiet repricing could spin a different narrative.