AI Trade Cools as Tech Lags Broader Market on Capex Payoff Doubts (NVDA, AVGO, MSFT)

The relentless AI rally that powered markets through 2025 is showing fatigue. After briefly retaking sector leadership in late May, technology is again lagging the broader tape as investors interrogate the central question of the cycle: will the staggering capital expenditure actually pay off in the near term? The numbers behind the unease are enormous. The five biggest spenders—Amazon, Alphabet, Meta, Microsoft and Oracle—are on track to commit roughly $725 billion to AI infrastructure in 2026, up about 77% from 2025's record of around $410 billion. The Big Four hyperscalers are expected to plow some 94% of operating cash flow into capex this year, versus a 10-year average near 40%. Alphabet's recent $80 billion equity issuance underscored how the spending bar keeps rising. The revenue side has not kept pace. Sequoia's David Cahn pegs the annual AI revenue gap that must be filled to justify the buildout at roughly $600 billion—and that gap is widening, not closing. Microsoft's targeted $25 billion in AI-related revenue for fiscal 2026 sits against capex estimates ranging from $97.7 billion to $150 billion, implying payback periods stretched over many years. As one analyst framed it, the market is no longer satisfied with the promise of AI; it wants demand-side cash flows. The catalyst this week was Broadcom (AVGO). Its shares tumbled double digits after it failed to raise its AI semiconductor sales forecast for 2026, dragging chip peers Nvidia (NVDA), Marvell (MRVL) and AMD lower in premarket trading. The episode stress-tests the AI capex story at the customer level for the first time, rather than at the supplier level. Market breadth tells the rotation story. The equal-weight S&P 500 has outpaced the cap-weighted index year-to-date, signaling gains are coming from the average stock rather than a handful of mega-caps. Over the past month, the so-called Magnificent Seven slipped slightly even as the index pushed above 7,600 for the first time. Concentration risk looms large: semiconductors now make up roughly 17% of the entire S&P 500, and five chip names are worth more combined than Apple and Microsoft together. Not everyone is bearish. BlackRock's Rick Rieder says he feels "a bit more relaxed" than during the dotcom era, and Goldman Sachs still sees the S&P 500 rallying about 12% in 2026. Many strategists characterize the slowdown in token consumption as a J-curve "digestion phase" rather than terminal demand weakness. Still, the message from the tape is clear: with valuations stretched and the buildout fully understood, the market will no longer pay any price for AI exposure. Until the demand-side cash flows arrive, the AI trade looks set for a pause.