Quick Facts
- 86% of enterprises running their own GPUs report utilization of 50% or less, per a June 2026 VentureBeat Research survey of 573 technical leaders.
- Average GPU utilization across 23,000 enterprise Kubernetes clusters sits at just 5%, according to Cast AI’s 2026 State of Kubernetes Optimization Report.
- Hyperscalers are on pace to spend $700 billion on AI infrastructure in 2026, nearly double 2025 levels, while enterprises report widespread idle capacity.
Enterprise GPU capacity is sitting largely idle. A June 2026 VentureBeat Research survey of 573 technical leaders at companies with 100 or more employees found that 86% of enterprises running their own GPUs report utilization of 50% or less. Of those surveyed, 81% said they recommend or decide AI purchases at their companies.
The gap between deployed capacity and actual use runs deeper than self-reported figures suggest. Cast AI’s 2026 State of Kubernetes Optimization Report, drawn from measured production telemetry across 23,000 clusters, puts average GPU utilization at just 5%. That means 95% of provisioned GPU capacity is idle at any given moment.
Cast AI co-founder and President Laurent Gil told VentureBeat that 5% is about six times worse than a no-effort baseline. He described many of the newer specialized compute providers as “neo-real estate” rather than true cloud infrastructure.
The Agent Gap
Enterprises are also overstating their AI agent deployments. The survey found that 71% of enterprises say a quarter or fewer of their deployed agents can complete multi-step work on their own. Only 10% say true autonomous agents make up the majority of what they run.
Gartner has flagged this pattern as “agentwashing” — vendors and companies labeling basic AI assistants as agents. In a June 2025 press release, Gartner estimated only about 130 of the thousands of vendors claiming to offer agentic AI are actually delivering it. The research firm projects 40% of enterprise applications will be integrated with task-specific AI agents by the end of 2026, up from less than 5% in 2025.
Anushree Verma, Sr. Director Analyst at Gartner, said AI agents will move from task-specific tools toward agentic ecosystems, shifting enterprise applications from individual productivity tools into platforms for autonomous collaboration.
Governance and Cost Controls Are Missing
Only 44% of enterprises rigorously track what their AI compute costs and returns. The rest are estimating. Twenty-seven percent exercise only reactive control over agent spend, learning what an agent costs when the invoice arrives, with no per-agent budget or ceiling in place.
Security gaps are also widespread. Fifty-four percent of companies had an agent security incident or near-miss in the past 12 months. Sixty-nine percent run agents with credential sharing somewhere in their deployments.
Sid Nag, President and Chief Research Officer at Tekonyx, said the real bottleneck is not model capability. “It’s the lack of production-grade architecture, data readiness, and operating models needed to turn AI from experimentation into enterprise-wide systems,” he said. Nag put typical utilization in Kubernetes-based AI clusters at 15% to 25%, still well below capacity.
The Spending Mismatch
The enterprise underutilization data lands as hyperscaler spending climbs sharply. Microsoft, Alphabet, Amazon, Meta, and Oracle have collectively committed to between $660 billion and $690 billion in capital expenditure for 2026, nearly double 2025 levels. Hyperscaler capex is now consuming close to 90% of operating cash flow, according to Barclays.
Gartner estimates AI infrastructure is adding $401 billion in new spending this year. Yet the enterprise data shows organizations have not yet built the architecture, data governance, or operational controls to run what they already own at meaningful scale. Only 25% of enterprises run a governed semantic layer in production. Forty-one percent have not started building one.
Vendors are watching. Forty-five percent of enterprises say an AI-specialized cloud such as CoreWeave, Lambda, or Crusoe is the compute option they are most likely to evaluate in the next 12 months. Today, fewer than 2% of those same enterprises report using one. Roughly six in 10 enterprises plan to switch or add vendors across key control layers within the next year.
This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.