Wall Street's forecasts for Big Tech require a leap of faith: that AI hyperscalers can boost revenue much faster than costs. Investors are uneasy about soaring capital expenditures on artificial intelligence. S&P cut Oracle's credit rating near junk, and Alphabet shares fell after raising capex estimates. Microsoft, Meta Platforms, and Amazon.com report earnings this week.

The bullish narrative assumes newfound operating efficiency amid trillions in capex. Consensus estimates for five hyperscalers project operating margins improving to 31% in 2029 from 27% last year—the highest since Meta's 2012 IPO, per Visible Alpha data. Half the improvement relies on SG&A falling to 8% of revenue from 10%, while revenue nearly doubles to over $3 trillion.

This leaves little error margin. Depreciation will soar from new buildings and equipment. A two-percentage-point SG&A drop compounds rapidly: at $3 trillion revenue, the five would spend $77 billion less on operations in 2029 than if overhead stayed at 2025 levels. That efficiency squeeze drives margin forecasts.