In recent months, several AI service providers — including Anthropic, OpenAI and GitHub — have moved parts of their offerings from flat-fee subscriptions to usage-based billing. That shift has added new uncertainty for companies trying to project, track and manage operational AI costs as they scale deployments.
Key findings from the KPMG survey
KPMG surveyed 2,145 senior executives across 20 countries. Main findings include:
- 29 percent of respondents find it difficult to understand and control the operational costs associated with enterprise AI as deployments scale.
- About one-third reported limited knowledge of AI costs and economic implications, complicating the rollout of AI agents.
- Nearly half of organisations rescheduled AI projects when costs exceeded the expected business value.
- The fastest-growing influence on AI strategy is the availability of lower-cost yet high-accuracy models — their importance rose by 7 percentage points compared with the previous quarter.
KPMG interprets these actions not as a decline in confidence in AI, but as a growing willingness to test where AI actually creates value and to focus investments where expected returns are strongest.
How companies are responding to changing cost structures
Usage-based pricing and rising fees have prompted firms to rethink AI strategies and, in some cases, delay or downscale projects when costs outstrip expected benefits. At the same time, large cloud providers are significantly increasing capital spending to expand AI infrastructure.
According to the report, Amazon plans roughly $200 billion in capital expenditure this year — a 50 percent increase over the prior year — much of which is directed at building AI capacity in AWS data centers. Microsoft’s total capital expenditure is expected to reach about $190 billion, a 61 percent increase year-on-year.
Both companies are also investing heavily in customer-facing engineering teams to accelerate adoption of AI applications: Amazon announced a $1 billion investment to create an AWS Forward Deployed Engineering organisation aimed at helping customers deploy AI agents more quickly; Microsoft has committed $2.5 billion to a new operating unit, the Microsoft Frontier Company, intended to help customers amplify their intelligence with AI while refining value propositions in their markets.
Governance and accountability remain major issues
The KPMG report highlights AI governance as a continuing challenge — namely, defining who is accountable for decisions produced by statistical models that can sometimes produce incorrect outputs, so-called "hallucinations." KPMG stresses that leadership accountability is important, but the ultimate success or failure of governance hinges on day-to-day operational practices.
The report recommends clear rules on issues such as when employees can intervene, who bears AI-related costs, how AI outputs are reviewed, and what happens when systems fail. While most organisations say they have some governance mechanisms in place, relatively few assert that these practices are fully embedded into daily operations.
Issues with KPMG’s own report and subsequent actions
KPMG’s report titled "Total Experience: Redefining Excellence in the Age of Agentic AI," published in October 2025, came under scrutiny after researcher GPTZero said it found only five out of 45 citations that accurately pointed to the referenced source; the remainder contained various errors ranging from misleading or fabricated details to overly vague, unverifiable references. Following that review, KPMG removed the report from certain websites and initiated a review of the circumstances that led to its publication.
A KPMG spokesperson said the firm places high importance on the accuracy and credibility of published content, and that it removed the report pending a review. The spokesperson reiterated KPMG’s responsible AI usage policies, which include human oversight of content and independent verification of sources.
Conclusion
The shift toward usage-based pricing, rising infrastructure investments by major cloud providers, and the challenges of cost transparency are reshaping corporate AI strategies. KPMG’s survey indicates many executives feel uncertain about managing AI expenditures, leading some organisations to delay or adjust projects while prioritising investments with clearer expected returns. At the same time, the episode around KPMG’s own report underlines the importance of rigorous sourcing and governance — both in research and in operational AI deployments.
