Large U.S. companies are reassessing their AI investments. Two AI company executives interviewed by CNBC stated that as the cost of model deployment continues to rise, many finance managers are redirecting budgets that might have been used for hiring expansion towards AI spending, and companies are paying closer attention to whether each model expenditure truly yields a return.
Annual budget to be exhausted in a few months
Glean CEO Arvind Jain stated that the biggest concern for many companies right now is the ballooning AI budgets. Some companies that originally allocated their AI budgets for the entire year may find them nearly exhausted within one or two months.
He stated that costs have not decreased as buyers expected; instead, they have continued to rise with the rollout of new-generation cutting-edge models. These new, token-denominated models are often more expensive than their predecessors, rapidly amplifying the cost pressures on companies deploying AI.
Jain believes this shift is allowing companies to directly compare technology costs alongside human resource costs for the first time. Previously, technology spending typically accounted for only a small portion of operating costs, but now management is discussing whether to continue increasing staff or allocate more budget to AI systems.
High-cost models are still used for simple tasks.
A major source of cost pressure is the inefficiency of enterprises in choosing models. Jain stated that approximately 95% of enterprise AI usage still relies on the most expensive cutting-edge models, even though many of these tasks could actually be accomplished by cheaper models.
He believes the most direct way to reduce costs is to offload simple tasks to lower-cost models. By optimizing model routing at the task entry point, businesses can achieve significant cost reductions, with savings of up to 10 times.
Factory AI CEO Matan Grinberg expressed a similar view. He stated that over the past year, companies' attitudes toward AI have gone through three phases: first, boards of directors demanded that management push AI forward as quickly as possible; then, they entered a phase of "using it regardless of cost"; and now, they are beginning to re-examine which tasks truly require top-of-the-line models.
Buyers begin to weigh prices
Grinberg says that corporate management is now facing a clear resource allocation problem: whether to optimize the number of employees or optimize AI spending per employee. This means that AI budgets are no longer just new investments, but are beginning to affect broader human resources and operational arrangements.
Both interviewees believed that the problem with AI is not that it is unusable, but rather that there is still a gap between "usability" and "cost-effectiveness" at this stage. The technology has already demonstrated its value, but the returns for businesses have not yet fully covered the continuously rising costs of its use.
This also presents new challenges to the current AI narrative in the market. If large enterprise clients become increasingly price-sensitive, business models that rely on charging for high-priced models will face greater pressure, and the growth expectations for the AI industry may need to be reassessed.
