Hyperscalers' $725B AI bet faces user rejection as trust collapses

A stark disconnect is emerging in the AI industry: hyperscalers are collectively committing roughly $725 billion in AI capex for 2026—about 75% of it AI-specific—while the very users they target are actively rejecting AI-generated output. Gartner's 2026 survey finds 50% of US consumers prefer brands that don't use generative AI, and 58% trust brands less for using it. Only 15% trust them more, and 54% report AI fatigue. This isn't just sentiment; it translates into measurable behavior. Wikipedia's English editors voted 44-2 to ban AI-generated content across 7.1M articles. Stack Overflow's new-question volume plunged 78% year-over-year, from a peak of 200,000/month in 2014 to just 3,862 in December 2025. Google AI Overviews have driven a 58% collapse in click-through rate for top search results, with Ahrefs documenting the decline doubling in 12 months. Chartbeat data shows Google referral traffic down 33% globally across 2,500 news sites.

This user rejection is not just a cultural commentary—it is enabling a fraud economy. Haotian AI ran a deepfake ring cloning voices and faces on WhatsApp, Zoom, and Teams, scamming victims out of $4M. Meanwhile, the industry's financial structure is showing strain: SoftBank cut its OpenAI margin loan target 40% to $6B, and Meta dropped 7% on $145B capex guidance—the harshest market reaction this cycle. Even consumer brand CEOs are stepping down citing AI uncertainty—Coca-Cola's James Quincey and Walmart's Doug McMillon both told CNBC that navigating the next wave of AI was a key reason for their exits. The market is adding capacity fastest in the exact areas where buyers are walking away, creating a structural tension that now defines every major AI story.