A survey by VentureBeat Intelligence reveals that 79% of 218 enterprises currently using AI in production report that their GPUs remain idle at least half of the time. Data collected between July and September indicates that 31% of these respondents utilize their hardware for productive tasks only 25% of the time or less.
Despite low utilization rates, 70% of companies reporting underused hardware intend to adopt new, additional, or replacement AI compute platforms within the next 12 months. In the September survey wave, 63% of all qualified respondents expressed a similar intent to switch or expand their current infrastructure.
Financial oversight remains limited, as only 10% of the surveyed enterprises have a finance-reviewed return on investment for their existing AI infrastructure. Additionally, just 12% of the companies reported utilizing specialized AI cloud providers, such as CoreWeave or Lambda.
The research suggests that the discrepancy between low hardware usage and continued investment may create opportunities for tools designed to optimize existing deployments. Analysts note that without formal ROI assessments, current spending patterns are driven more by strategic conviction than by measured performance metrics.
