UBS: The rising cost of enterprise AI adoption is due to the surge in usage, and the market overestimates the inflation risk of tokens According to BlockBeats, on June 15th, UBS (America) stated in its latest research report that enterprise AI adoption is facing new frictions caused by rapidly rising token and computing power costs, but this problem is more due to the surge in usage rather than unit price inflation, and overall risk may be overvalued by the market. The report points out that with the deployment of high-intensity tools such as AI encoding agents, the consumption of enterprise tokens far exceeds expectations. This phenomenon has frequently appeared in investor discussions and has raised concerns about the possible slowdown in the diffusion speed of AI technology on the enterprise side. UBS found through interviews with IT executives from approximately 13 companies that about 60% of the surveyed organizations have identified AI token and computing power costs as substantive issues, especially after transitioning from simple chatbots to autonomous proxy applications, where costs have shifted from fixed SaaS expenses to variable consumer expenses and budget predictability has significantly decreased. Most companies have already or plan to introduce barrier measures, including token pooling, model downgrade usage, waste reminders, and restrictions on heavy users, to eliminate obvious waste rather than comprehensively curb adoption. Some executives have made it clear that they are unwilling to significantly restrict employees' use of AI, emphasizing that "our goal is to get employees to start using AI." Therefore, they have chosen to optimize other budgets by cutting external IT services, integrating cloud spending, and other means to balance the rising cost of AI. The report emphasizes that almost all surveyed companies mentioned that AI adoption rates are accelerating, especially among developer teams, indicating that cost increases are mainly driven by usage growth rather than unit cost inflation. UBS believes that this situation is a normal cost control behavior of enterprises, not a signal of AI adoption obstruction. Even companies like Uber, which have already used up their annual AI budget within a quarter, still maintain high token limits and fully promote AI applications, while hedging costs by improving engineer efficiency. UBS further analyzed that AI model providers and ultra large scale cloud service providers are accelerating token efficiency improvements, which may limit recent price increases and have an impact on the distribution of cloud service provider shares. Google Cloud and AWS may gain advantages in cost control through self-developed chips and models. At the same time, the resistance of enterprises to the usage pricing model may increase, or lead to further pressure on the non AI software expenditure environment. Based on the previous 140 enterprise AI survey, the report points out that "unclear investment returns" are still the biggest obstacle to adoption, and "lack of budget" has not yet entered the top five. However, as token cost issues become more prominent, this dynamic is becoming a key factor for enterprises to optimize AI deployment more pragmatically.