Enterprise GPU Utilization at Half Capacity or Less, Study Finds

A significant portion of enterprise-owned high-performance hardware, such as GPUs, is underutilized, according to a June survey by VentureBeat Research. The study surveyed 573 technical leaders at companies with 100 or more employees.

Expensive Hardware Sits Idle

The findings indicate that 86% of enterprises running their own GPUs report utilization rates of 50% or less. This data comes directly from the companies making the AI investments and adds a layer to the ongoing debate on Wall Street regarding the valuation of AI infrastructure.

Lack of Control and Rising Costs

Many companies are facing challenges with AI agent implementation. 54% of businesses have experienced security incidents or near-misses related to AI agents that were caught before causing harm. Furthermore, 27% of companies only track AI spending reactively, learning costs only after invoices arrive without per-agent budgets.

Businesses are planning to improve AI agent oversight. Roughly six in ten enterprises intend to switch or add vendors across five control layers within the next 12 months. These layers include agent identity, output evaluation, cost telemetry, context, and orchestration.

Actual Agent Capabilities

The survey also highlighted that most deployed AI agents are performing simple, single-prompt tasks. 71% of enterprises stated that a quarter or fewer of their agents can complete multi-step tasks autonomously. Only 10% reported that true agents constitute the majority of their AI deployments.