Enterprises Are Spending Billions on AI Compute They Cannot Track or Justify

Quick Facts

• 83% of enterprises report GPU utilization at 50% or below, with 49% running at 25% or below, per VentureBeat Pulse Research.
• Only 22% of finance executives can tie AI spend to business outcomes, even as enterprise AI budgets average $85,521 per month in 2025.
• 64% of enterprises plan to switch or add an AI infrastructure provider within 12 months, signaling unusual churn in a foundational spending category.

Enterprises are spending aggressively on AI infrastructure and have almost no clear view of what they are getting for it. A VentureBeat Pulse Research survey of 107 enterprise respondents, fielded in June 2026, finds GPU fleets sitting largely idle while budgets climb and finance teams struggle to connect dollars spent to business results.

The report calls this the compute gap: capital moving faster than the controls needed to manage it.

GPU Fleets Are Largely Idle

The utilization numbers are striking. Eighty-three percent of enterprises operating GPUs report utilization at 50% or below. Nearly half, 49%, run at 25% or below. Only 12% clear the 50% threshold. Eight percent do not measure utilization at all.

An independent dataset from Cast AI’s 2026 State of Kubernetes Optimization Report puts actual production cluster utilization at roughly 5%. The survey numbers are bad. The real-world numbers may be worse.

Despite this, enterprises plan to buy more GPU capacity. They are adding to fleets that are already substantially underused.

The Bill Surprises After Launch

Enterprise AI budgets are averaging $85,521 per month in 2025, a 36% increase from 2024. Gartner forecasts worldwide AI spending will reach $2.59 trillion in 2026, up 47% year over year. Enterprise generative AI spending jumped from $11.5 billion in 2024 to $37 billion in 2025, according to Menlo Ventures.

Yet only 51% of organizations say they can confidently evaluate whether those investments are delivering returns. Between 30% and 50% of AI-related cloud spend goes to idle resources, overprovisioned infrastructure, and poorly optimized workloads.

J.R. Storment, Executive Director of the FinOps Foundation, told TechCrunch that companies started calling in a panic early this year. “In April and May, I started hearing from companies: ‘Oh my god, we are 3x over our entire 2026 token budget and it’s only April,'” he said. “We started hearing existential crises, and the whole conversation shifted from tokenmaxxing and ‘go fast’ to ‘we need guardrails, how do we control this?'”

Uber burned through its entire 2026 AI coding budget by April. Microsoft revoked developer Claude Code licenses months after enabling them. A Priceline employee reported a routine Cursor contract renewal came back four to five times more expensive than expected.

ROI Remains Hard to Prove

An MIT Project NANDA study from July 2025 found that 95% of enterprise generative AI pilots produced zero measurable profit-and-loss impact on roughly $30 billion to $40 billion in corporate spending. Conditions have improved since then, but the lag between infrastructure buildout and genuine return on investment remains wide.

Nishant Gupta, Chief Availability Officer at Salesforce, described the underlying problem. “Token economics is fundamentally more abstract and opaque than anything we’ve managed at this scale before,” he said. “It requires a different operational muscle than the one the industry built for cloud.”

Only 22% of finance executives can connect AI spend to business outcomes, according to the survey. That number explains why GPU spend has become a board-level issue. Organizations reporting AI as an active financial operations concern jumped from 31% in 2024 to 63% in 2025, per CloudZero.

Provider Loyalty Is Fragile

Sixty-four percent of enterprises plan to switch or add an infrastructure provider within 12 months. Thirty-eight percent plan to do so within the next quarter. For a spending category this foundational, that churn rate is high.

When choosing providers, enterprises prioritize integration with existing systems, cited by 41%, and total cost of ownership, cited by 35%. Headline token price is the deciding factor for just 8%.

One emerging constraint is barely registering. The shift from GPU compute to memory bandwidth as inference scales is a factor for roughly one in five enterprises, with the rest either unaware of it or not yet addressing it.

What Comes Next

IDC Research Director Juan Seminara described the spending pattern as structural, not cyclical. “The fact that spending accelerated throughout the year, reaching nearly $90 billion in the final quarter alone, tells us that enterprises and hyperscalers are not building for today’s workloads, but for AI architectures that are still being defined,” he said.

Global AI infrastructure spending is forecast to exceed $1 trillion by 2029, growing at a compound annual rate of roughly 31% from 2025. The companies that build cost visibility and utilization discipline now will have a material advantage over those still flying blind when that bill comes due.

Read more: The AI compute gap: Enterprises are buying infrastructure faster than they can measure what it costs