Enterprises Prioritize AI Compute Speed Over Cost Visibility

A recent survey reveals that two-thirds of enterprises have AI workloads running in production, with nearly 30% operating them at scale. Concurrently, companies' ability to accurately track the costs associated with their AI compute infrastructure has not kept pace.

Traditionally, cost has been a primary driver in purchasing decisions. However, for AI, this is shifting. Speed and availability have become more critical than total cost of ownership (TCO), and reliability is now prioritized over price as a success metric. This reordering of priorities likely stems from the pressure to deploy AI solutions effectively in production environments, but it highlights a challenge: fewer than half of businesses can rigorously track their AI compute expenses.

Most deployed GPU accelerators are operating at 50% capacity or less, and future investments are targeting specialized clouds that currently see limited adoption. Leading platforms include OpenAI, Google Gemini, and Microsoft Azure, with Azure serving as the primary platform for 26% of respondents.

Planned investments show a strong interest in AI-specialized cloud environments, despite low current usage. Non-Nvidia accelerators are also gaining traction. Many enterprises intend to add or switch providers within the next year, though current major players often dominate these considerations.

The research, which surveyed 170 enterprises, focused on AI infrastructure, compute, and economic factors. The findings indicate a move towards production-level AI deployment but underscore the need for improved cost management and visibility.