The Magnificent Seven are planning to spend over $680 billion on capital expenditures in 2026. Most of it goes to AI infrastructure. That figure is roughly 70% higher than last year's already historic levels. And the market's response? A collective shrug, a handful of analyst upgrades, and the quiet assumption that all of this will work out.
It won't all work out. It never does. The last time capital spending scaled this aggressively relative to expected returns was 1999. We know how that ended.
The Numbers That Should Keep You Up at Night
Let's start with the spending. Amazon has committed to $200 billion in capex for 2026. Alphabet is guiding to $175–$185 billion, nearly double what the company spent last year and $60 billion above what Wall Street analysts had expected. Microsoft is tracking toward $144 billion. Meta is projecting $115–$135 billion. Oracle, the most leveraged player in this game, is targeting $50 billion while its five-year credit default swap has more than tripled since September.
Follow the money, not the narrative. Pivotal Research projects that Alphabet's free cash flow will plummet almost 90% this year, from $73.3 billion to $8.2 billion. Amazon's trailing-twelve-month free cash flow sits at just $11.2 billion against that $200 billion capex commitment. That means Amazon expects to be deeply cash-flow negative in 2026. These are not the financial profiles of companies making prudent investments. These are the profiles of companies engaged in an arms race where the penalty for losing is existential, so they spend as though the penalty for overspending doesn't exist.
Bank of America estimates that AI capex among the mega-cap tech firms will consume 94% of operating cash flows (minus dividends and buybacks) in 2025 and 2026, up from 76% in 2024. J.P. Morgan estimates that data center construction alone will require $1.5 trillion in investment-grade bonds over the next five years. Morgan Stanley puts data-center-related debt at over $1 trillion by 2028. These are staggering numbers, and they demand a simple question: where are the returns?
95% Failure, 100% Confidence
An MIT Media Lab study published in August 2025 examined 300 public AI initiatives and found that 95% of organizations are getting zero measurable return on their generative AI investments. Zero. Despite $30–$40 billion in enterprise spending. Only 5% of custom enterprise AI tools even reach production deployment. The study attributed the failures to brittle workflows, lack of contextual learning, and fundamental misalignment with day-to-day operations.
This isn't some fringe bear case. This is MIT. And the findings are consistent with what the consumer market is telling us. PCWorld documented that Dell relaunched its XPS brand at CES 2026 with AI scrubbed from the marketing. Back to longevity, everyday performance, lightweight design. The things people actually buy laptops for. If the end users don't want the product, who exactly is generating the revenue to justify $680 billion in spending?
Marcus laid out the bull case well in his recent work on Big Tech earnings, and the revenue growth at companies like Meta is genuine. But the bull case assumes nothing breaks under the surface. And right now, I keep finding things that might. The Case-Shiller price-to-earnings ratio for the U.S. market has exceeded 40 for the first time since the dot-com crash. The S&P 500's equity risk premium has collapsed to 0.02%, meaning investors are accepting equity volatility with essentially zero compensation above Treasuries. In late 2025, 30% of the S&P 500 and 20% of the MSCI World index was held by just five companies. That's the greatest concentration in half a century.
The Catalyst Nobody Is Pricing In
When everyone's bullish, I get nervous. A December Bank of America survey found that investors themselves see an AI bubble as the biggest tail risk event. More than half of respondents identified the Magnificent Seven as Wall Street's most crowded trade. And yet positioning hasn't changed. The crowd acknowledges the risk and then does nothing about it. That's not rational analysis. That's addiction.
Here's the specific trigger I'm watching. The hyperscalers are increasing capex by more than 60% from 2025 levels. If Q2 and Q3 earnings reports show that this massive acceleration in spending is not producing a corresponding acceleration in AI-driven revenue, the narrative cracks. It doesn't need to shatter; it just needs to crack. A single disappointing quarter from Nvidia, a slowdown in cloud backlog growth at Microsoft, a missed revenue number at Amazon Web Services. Any one of these would force a repricing of the entire AI supply chain, from chip suppliers to data center REITs to the power utilities that have been bid up on AI demand projections.
The Schwab U.S. Dividend Equity ETF, which owns none of the hottest AI tickers, has been quietly outperforming while AI stocks have flattened. Since October 2025, gold, biotech, energy, and materials have led gains while Nvidia, Tesla, Meta, and Microsoft have lagged sharply. The smart money rotation may already be underway. The equal-weight S&P 500 is beginning to outperform its cap-weighted counterpart for the first time in the AI era.
Deutsche Bank research suggests that if you strip out Big Tech capex, the U.S. economy would almost certainly be in recession. AI-related investment now accounts for roughly half of GDP growth. A reversal wouldn't just hurt tech stocks. It would risk recession. That's not diversification. That's dependency.
Complacency has a cost. And the bill for $680 billion in faith-based capital allocation is coming due. The question isn't whether the AI bubble pops. It's whether you've positioned for the possibility that it does, or whether you're still holding your breath and hoping the math works out.
