Report: Companies are buying AI computing power faster than they can track costs

In short: Big companies are increasing spending on AI computing, but many still cannot clearly measure and manage the total cost.

What's going on

A report looking at 107 enterprises says spending on AI infrastructure is rising quickly, faster than many teams can track what it really costs. AI infrastructure is the computers, chips, and cloud services used to run AI systems.

Most companies in the report still run AI using familiar large cloud providers and AI model APIs. An API is like a remote control that lets a company use an AI model hosted by someone else. But the next wave of spending is aimed at more specialized computing hardware, which many of these organizations do not use today.

The report also says many companies expect to change their setup soon. A majority plan to switch providers, or add a new provider, within the next year. Many expect to do it even sooner, within a quarter.

When companies decide what to buy, the report suggests they are not focusing only on the posted “per token” price, meaning the price per small piece of text the AI reads or writes. Instead, they care more about how well the new tools fit into their existing systems and the total cost of owning and running them over time. You can think of it like choosing a delivery van, where the sticker price matters, but so do maintenance, fuel, and whether it works with your routes.

What to watch

Look for more tools and services that help companies measure AI costs across different providers, especially as more teams mix cloud services with specialized hardware. The companies that can track costs clearly may be able to expand AI use faster, while avoiding surprise bills.

Source: Venturebeat