Forty-five percent. That is the share of Microsoft's $625 billion cloud backlog tied to a single customer: OpenAI. A company that burns cash, depends on continued fundraising, and just signed a $38 billion deal with Microsoft's competitor AWS. Strip OpenAI out, and Microsoft's remaining backlog grew 28%. Solid, but not the 110% headline number that made the press release. Nobody is talking about this.
But this column is not about Microsoft specifically. It is about a market structure that mechanically funnels capital into a handful of companies whose internal risk profiles are far more fragile than their index weightings suggest. And about the trillion-dollar passive flow machine that makes that structure self-reinforcing.
The Reflexive Loop Has a Name: It's Called Your 401(k)
You already know the shared facts. Passive holds 55% of fund assets. A trillion-dollar net swing from active to passive in a single year. The top 10 stocks are 39% of the S&P 500. The Shiller CAPE briefly touched 40, a level reached only once before, during the dot-com peak. These are not opinions. They are coordinates on a map, and the map says we are in unusual territory.
The standard rebuttal is that active investors still drive 95% of daily volume, so price discovery is fine. I have heard this argument for years. It is technically true and practically misleading. When passive funds absorb $951 billion in a year and active loses $187 billion, the marginal price-setter changes even if the volume numbers don't. Valentin Haddad's research found markets are already 11% less elastic because of passive growth. That was measured in 2021, when passive held roughly 40% of fund assets. It holds 55% now.
What does reduced elasticity mean in practice? Prices move less in response to new information on the way up. They can move a lot more on the way down, because the remaining active participants who should be buying the dip are a shrinking pool. Research Affiliates put it plainly in their 2025 paper: passive strategies "cause unrelated stocks to move synchronously, undermining diversification." When everyone owns the same basket, correlation spikes exactly when you need it not to.
Follow the Money Inside the Money
The earnings bulls point to Mag 7 profit growth and say concentration is justified. Fine. But look one layer deeper.
Nvidia's four largest direct customers accounted for 61% of its fiscal Q3 2026 revenue, up from 36% a year earlier. That is customer concentration inside the most concentrated stock in the most concentrated index in the most passive market in history. It is concentration all the way down.
Microsoft just disclosed that 45% of its $625 billion commercial backlog comes from OpenAI, a company developing its own custom AI chips with Broadcom, with deployments starting late 2026. Microsoft lost $357 billion in market cap after reporting earnings that beat estimates. The market looked past the headline and saw that margins are compressing, not expanding, under the weight of $37.5 billion in quarterly capex.
Meanwhile, the four hyperscalers collectively plan to spend between $635 billion and $665 billion in 2026 capex, a roughly 67% spike from 2025. Pantheon Macroeconomics calculated that without AI spending, U.S. corporate capex would be negative. That is not a sign of broad economic health. That is one bet, funded by a handful of companies, carrying an entire investment category.
Alphabet warned that 2026 expenses will "meaningfully increase" due to depreciation and energy costs from data centers. Amazon's shares dropped 7% after announcing $200 billion in 2026 capex, blowing past analyst expectations of $146 billion. Apple, which spent $12 billion in annual capex and outsourced its AI to Google, was the only Mag 7 name to rally after earnings. The market is starting to distinguish between spending and earning. Passive flows do not make that distinction.
The Illusion of Diversification
Here is the part that should concern every index fund investor. The "Great Rotation" story gaining traction in early 2026, with 65.8% of S&P 500 stocks above their 200-day moving averages, is being presented as evidence that the market is broadening. And maybe it is. But the fact that we celebrate two-thirds of an index participating as a "critical milestone" tells you how narrow the market has been. In a healthy market, that number is unremarkable.
The real test of this rotation will come under stress. During the April 2025 tariff selloff, the percentage of S&P 500 stocks above their 50-day moving average collapsed to 4%. Four percent. That is what happens when passive mechanics meet a genuine shock. Breadth evaporated in days. It took weeks to rebuild.
And the VIX sits at roughly 15 as of late February 2026. The long-term average is 20. Sub-15 readings have historically preceded, not prevented, volatility spikes. When everyone is bullish, I get nervous.
The credit market is whispering what equities refuse to hear. In early February, $17.7 billion in tech loans dropped to distressed levels over four weeks, the most since October 2022. High yield spreads touched 275 basis points before bouncing. Oracle's five-year CDS has more than tripled since September. The companies borrowing to fund AI capex are starting to get priced differently by the people whose only job is to assess whether they'll get paid back. Credit markets are smarter than equity markets, because bond investors do not get upside.
The last time concentration was this extreme, the Shiller CAPE was this elevated, and a single technology narrative dominated capital allocation this thoroughly, the year was 1999. The comparison is not perfect. Today's companies are more profitable. But profitability does not immunize you from repricing when the assumptions baked into a 40x CAPE start to crack. A 50% loss requires a 100% gain to recover. That math does not care how good your earnings were last quarter.
Three things to watch: hyperscaler free cash flow conversion in Q3 and Q4 2026, the AI startup mortality rate as 2022-2023 vintage venture funding runs dry, and whether the passive flow machine slows if equity returns flatten. The infrastructure built today may take 18 to 36 months to generate proportional returns. If that timeline stretches, the gap between spending and earning becomes the story. And passive funds will mechanically own every dollar of that gap.
Position sizing. Hedging. Cash as a position. These are not signs of fear. They are signs of arithmetic.
