Companies rushing to adopt AI may soon face unexpected costs as many struggle to track how much they are actually spending on the technology, says a KPMG survey. The study found that only 26% of companies have a comprehensive view of their AI costs, while 50% have only partial visibility. Another 22% said they have no visibility into their AI spending or only see costs after receiving bills. Unlike traditional software subscriptions that often come with fixed fees, many AI services charge customers based on usage. Costs are measured through "tokens," which are units used to calculate how much computing power AI systems consume.According to Steve Chase, KPMG's global head of AI, many companies are struggling to keep up with the rapid growth in usage."It's a new resource that needs to be managed that didn't exist quite that way, and we're seeing exponential growth," Chase said as quoted in a report by The Wall Street Journal.KPMG said it is working with companies that exhausted their annual token and cloud-computing budgets within months. One client saw its token usage increase sixfold, according to Chase.CFOs face growing AI billsThe shift to usage-based pricing is forcing finance leaders to monitor AI consumption more closely. Gil Luria, head of technology research at D.A. Davidson, said the model places more responsibility on customers."Many other CFOs are going to see their Anthropic bill and freak out this quarter," Luria said.Companies including OpenAI, Anthropic, Microsoft and Salesforce charge enterprise customers at least partly based on how much they use AI services.Firms take measures to control spendingSeveral businesses are already taking steps to control spending. As per the WSJ report, Digital safety company Life360 is redesigning AI agents and using tools that reduce token consumption. Buy-now-pay-later company Affirm said its AI usage increased sharply after software teams began using AI agents to write more code.Amer Sports, the company behind brands such as Arc'teryx and Salomon, is also taking a measured approach."I didn't want everyone going out, you know, using tokens and setting up AI agents or setting up AI processes that didn't have long-term scalable benefits for the entire back-end process," finance chief Andrew Page said, per the report.