A new survey of 107 enterprises reveals a striking disconnect: organizations are pouring money into AI infrastructure faster than they can measure what it actually costs. The findings paint a picture of an industry in rapid expansion mode, but running blind.
Only 21% of enterprises currently run AI in production at scale, yet infrastructure spending intentions are racing ahead of that maturity. The single largest planned evaluation area over the next year is AI-specialized clouds (45%), a category almost none of these enterprises use today. This represents a fundamental re-platforming in the making.
The Utilization Problem
Perhaps the most revealing statistic: 83% of enterprises report GPU utilization at 50% or less. In practical terms, this means most organizations are sitting on expensive compute capacity they aren’t fully using. Fewer than half (44%) can rigorously track what their AI compute costs. The result is a compute gap—massive investment running ahead of the visibility needed to control it.
“When you buy infrastructure faster than you can measure its economics, you’re essentially flying blind,” noted one analyst familiar with the findings. “This isn’t a vendor problem—it’s a maturity problem.”
The Coming Switch Wave
The data also reveals remarkable instability in infrastructure choices. A full 64% of enterprises plan to switch or add an infrastructure provider within twelve months, with 38% planning changes within the next quarter. For a category this foundational, that’s extraordinary churn intent.
When they do switch, enterprises say integration (41%) and total cost of ownership (35%) drive decisions—not headline token pricing. Only 8% cite cost per million tokens as the deciding factor. This suggests enterprises are prioritizing flexibility and control over pure price competition.
What This Means
The findings suggest a multi-layered transition: from hyperscalers to specialized AI clouds, from single vendors to multi-provider strategies, and—critically—from reactive buying to more disciplined cost governance. Enterprises that master this transition first may gain significant competitive advantages in the AI era.
The question facing IT leaders isn’t whether to invest in AI infrastructure, but whether they can build the visibility needed to make those investments pay off.