Goldman Sachs says $700 billion in AI investment contributed “basically zero” to U.S. GDP growth in 2025, noting most AI equipment is imported and offsets domestic impact. Meanwhile, 70% of firms report using AI, but 80% see no productivity or employment gains.
Apptio reports nearly three-quarters of organizations are increasing IT spend, yet 90% of IT leaders now question AI ROI, up from 85% last year. Eighty percent cite data silos as a primary barrier to proving value.
Gartner predicts 40% of agentic AI projects will be canceled by 2027. S&P Global finds 42% are abandoned before production, often due to rushed deployments and weak infrastructure. Meanwhile, 84% report gross margin erosion from AI costs.
DigitalOcean reports 67% claim productivity gains from AI agents, yet only 10% have scaled them successfully. Forty-nine percent cite inference cost as the primary barrier. Adoption is rising; sustainable production isn’t.
Why do we care?
The $700 billion isn’t creating distributed value — it’s a margin transfer from enterprise IT budgets to a handful of infrastructure vendors. Your clients are on the paying end of that flow, and Goldman’s GDP finding confirms it at the macro level. The money is moving up the stack, not out into productivity.
The Apptio ROI concern jump — 85% to 90% in a single year — means internal IT budget owners are losing the justification argument inside their own organizations. You can’t defend spend you can’t measure, and you can’t measure across fragmented data environments. That’s a data infrastructure deficit that AI spending is actively making more expensive.
The burden of proof has inverted. AI must now recover margin before it can claim to generate value.
AI readiness is not model deployment. It’s data normalization, cost instrumentation, and governance discipline. MSPs positioning around AI features compete with hyperscalers. MSPs positioning around cost governance and data readiness build recurring revenue around a problem clients openly admit they have.