KPMG Survey Finds AI Usage-Based Pricing Is Making Enterprise Costs Harder to Predict

A KPMG survey of 2,145 executives across 20 countries found that companies scaling AI are struggling to forecast and manage costs as some major AI providers shift parts of their services from fixed subscriptions to usage-based pricing. According to ChainCatcher, the report cites recent moves by Anthropic, OpenAI, and GitHub toward pay-as-you-go billing models as a factor complicating enterprise budgeting for AI deployments. KPMG said 29% of executives reported difficulty understanding and controlling operating costs when deploying AI at scale, while about one-third said insufficient understanding of AI economics is hindering the rollout of AI agents. Nearly half of organizations, about 49%, have delayed or adjusted AI deployment plans because costs exceeded expected value. The report also said lower-cost, high-fidelity large models are increasingly influencing enterprise AI strategies. Separately, KPMG noted that large technology companies are increasing capital spending to build AI capacity. Amazon plans about $200 billion in capital expenditures this year and is investing $1 billion in AWS frontline engineering teams to help customers adopt AI agents. Microsoft’s total capital expenditures this year are expected to reach $190 billion, and it is providing $2.5 billion in funding for a new entity, Microsoft Frontier Company. Beyond cost pressures, KPMG said key challenges for enterprises include accountability in AI governance, rules for employee involvement, and preventing system hallucinations.