An October 2026 survey found that 56% of its qualified respondents had encountered an AI-related software charge or tier change in the previous 12 months, while 38% said cost overruns had disrupted an AI project. The results describe a self-selected group of respondents, not a market-wide estimate.
What respondents reported
VentureBeat Intelligence received 172 responses to its October survey and qualified 156. The survey covered people at organizations with at least 100 employees and consultants serving those organizations. Respondents self-selected.
| Survey measure | Result among qualified respondents | Period |
| Reported an AI surcharge, an AI-cited renewal increase, or existing features moved to a higher tier | 56% | Previous 12 months |
| Cited a cost overrun as a reason to delay, scale back, or cancel an AI project | 38% | Previous 12 months |
| Cited a computing-capacity shortage as a reason for AI project disruption | 12% | Previous 12 months |
The 56% figure combines three kinds of vendor change; it is not a measure of one product’s price increase. Among respondents who reported a vendor price or tier change, 40% said the added cost amounted to at least 10% of their organization’s annual software spending.
How broad access can reshape a projected budget
McKinsey describes one bank client whose projected annual AI spending rose from $3 million to more than $25 million after employees received unconstrained access to tools and models. It is an illustrative case involving one client, not a typical enterprise result.
For a rollout, the useful comparison is spending against work that succeeds. Track the cost of accepted tasks alongside retries and computing utilization, then assess returns using a financial or operational measure. A low price per token alone cannot tell you what a completed workflow costs.
What an AI charge measures
Token price, inference cost per workflow and cost per successful task describe different things. Gartner forecast token costs would fall 95% by 2030, while separately forecasting that inference costs per agentic workflow would rise more than fivefold through 2028. The Gartner forecast for agentic workflow inference costs concerns the cost of a workflow, not the price of an individual token.
In a May 2026 essay, Mark Smith argued that software pricing may shift from seats toward units of work, decisions or outcomes. He also warned that combining seat charges with usage fees can layer costs, while outcome-based pricing depends on measurement and governance.
