Analysis
While Nvidia, OpenAI and Meta chase hundreds of billions in AI infrastructure commitments, a quieter set of 2026 enterprise surveys tells a starkly different story about what that spending is actually buying. According to Writer's 2026 AI adoption survey, 56% of CEOs report zero measurable ROI from AI in the past 12 months, and only 12% have achieved both revenue gains and cost reductions from their AI investments.
The gap between adoption and value is widening, not closing: 79% of organizations report facing real challenges in adopting AI, a double-digit increase from 2025, and 54% of C-suite executives admit AI adoption is actively "tearing their company apart" organizationally. That's despite 97% of companies deploying AI agents in the past year and 52% of employees already using them day to day -- deployment has clearly outpaced value realization.
Separately, Publicis Sapient's 2026 Global Enterprise AI Report found AI now supports 53% of enterprise work, yet only 18% of organizations describe themselves as fully integrated. In the US specifically, 71% of respondents expect significant AI-scaling progress over the next 12 to 24 months, but only 20% say their organizations are actually equipped today to meet those expectations -- a striking confidence gap between where companies think they're headed and where they actually stand.
None of this is contradicted by rising budgets: 65% of enterprises increased their AI spending in 2026, with a median increase of 22% year-over-year, even as more than half of CEOs can't point to measurable returns. That combination -- rising spend, flat-to-negative measured ROI -- is exactly the dynamic investors should be watching as hyperscaler capex faces its own intensifying scrutiny this earnings season.
For GPs and operators, the practical read is that enterprise AI budgets are still expanding on faith and competitive anxiety rather than demonstrated returns, which means the vendors best positioned right now are the ones that can produce hard, auditable ROI numbers -- not the ones with the most impressive model benchmarks.
