The AI Computing Gap: Companies Are Buying Infrastructure Faster Than They Can Track the Costs

At a time when artificial intelligence has become the number-one strategic priority for most large organizations, a new report reviewed by VentureBeat shines a critical light on an increasingly apparent paradox: companies are spending record sums on AI infrastructure without having the tools to truly understand what they are buying — or what it is costing them.

Spending at full speed, visibility at an all-time low

The study, conducted across a sample of 107 companies, shows that AI infrastructure budgets are growing at a pace that far outstrips organizations' internal capacity to monitor and analyze those costs. Finance teams and technology teams alike find themselves in the paradoxical position of approving massive expenditures without access to clear metrics that could justify or optimize those investments.

According to VentureBeat, most organizations currently run their AI workloads on infrastructure provided by major cloud vendors — hyperscalers such as Amazon Web Services, Microsoft Azure, and Google Cloud — and through model providers' APIs. It is a familiar and relatively manageable approach, but one that is running into its limits as workloads continue to grow.

A paradigm shift: specialized infrastructure becomes the target

What is truly significant in the data analyzed is the direction in which future investment is heading. According to VentureBeat, "the next dollar is going toward specialized compute" — a category that almost none of the surveyed companies currently uses at scale. This includes computing solutions dedicated exclusively to AI workloads, such as AI-specific chips and data centers configured specifically for training and running large language models.

What is more, a majority of companies plan to switch or add infrastructure providers at some point this year — and a significant share intend to make that transition within the next quarter. This is a clear signal of a sector in full ferment, where decisions are made quickly, often under competitive pressure, and not necessarily on the basis of rigorous financial analysis.

The risk of decisions made in financial darkness

This dynamic raises serious questions for chief financial officers and those responsible for corporate strategy. How do you justify a major AI infrastructure investment to the board of directors when you lack the tools to measure the return on that investment? How do you optimize costs when you have no visibility into how computing resources are actually being consumed?

The gap identified by the report — between the pace of acquisition and the capacity to measure — is not merely an accounting problem. It is a strategic vulnerability. Companies that fail to resolve this equation risk accumulating costs that spiral out of control and losing the very competitive advantage they are trying to build through AI adoption.

What comes next for the AI infrastructure market

The trend toward multiple vendors and specialized infrastructure suggests that the market is in the midst of a maturation process. Companies are beginning to understand that generic cloud solutions are not always the most efficient option for compute-intensive AI workloads, and they are seeking alternatives more precisely calibrated to their specific needs. This is a natural evolution — but one that brings its own governance and financial control challenges.

For businesses across Europe and beyond, the lesson is clear: adopting AI does not simply mean buying technology. It means simultaneously building the capacity to measure, govern, and optimize it. Without that, the race for AI risks becoming a costly sprint with no visible finish line.