Wall Street's AI Bet Is Really a Bet on Cloud Operators

Amazon's strong second-quarter results sent its stock soaring nearly 10% after hours, driven by 20% net sales growth and exceptional cloud revenue. Yet what's most striking about the market's response is what it reveals about investor sentiment: they're increasingly confident in infrastructure builders but nervous about whether the AI business they're serving is economically viable.

The Infrastructure Spending Boom#

Amazon is doubling down on data center expansion despite warnings about rampant capex. The company spent $173 billion on property and equipment—covering GPUs, power systems, and land—in fiscal 2025, almost double the prior year's $107.65 billion. The company also raised its 2026 capex outlook to $220 billion, a $20 billion increase that came even as Amazon dipped into cash reserves for the first time this year, ending the quarter with $7.6 billion less cash than a year prior.

Ordinarily, such aggressive spending and negative free cash flow would spook shareholders. But Amazon's AWS cloud division provides crucial cover: it generated $42 billion in quarterly revenue, representing 37% year-over-year growth. That expanding revenue stream, paired with multi-year lags between construction and commercial operation, gives investors confidence the company knows what it's doing.

A Narrowing Definition of Success#

The pattern extends beyond Amazon. Microsoft and Google posted stock gains after reporting robust cloud businesses, while Meta—which is pouring capital into infrastructure without clear revenue offsets—saw its shares fall 8% amid investor concerns about cash flow and ongoing expenses.

The message is clear: Wall Street distinguishes sharply between companies that can invoice for their AI infrastructure and those that cannot. Amazon is hedging further by investing in custom chips like Trainium and Graviton processors, which don't appear as capex but promise improved margins that could justify the spending pattern.

The Fragile Foundation#

Yet this confidence rests on a fragile assumption. Cloud revenue is ultimately derived from what AI companies and their customers are willing to pay—and , meaning cloud hosts are simply collecting pass-through fees on larger bets about AI's economic value.