UBS: Rising Enterprise AI Adoption Costs Stem from Surge in Usage, Market Overestimates Token Inflation Risk

June 15 – In a new research report, UBS notes that enterprise AI adoption is facing new headwinds driven by sharp rises in token and computing power costs. However, the firm clarifies this challenge stems from a surge in usage rather than price inflation, broader market risk, or valuation overhang. The report highlights that deployment of high-intensity tools like AI coding agents is pushing enterprise token consumption far beyond initial forecasts, a trend that has come up frequently in investor discussions and stoked concerns over a potential slowdown in how quickly enterprises are integrating AI technology. UBS’s interviews with roughly 13 enterprise IT executives found that about 60% of surveyed organizations view AI token and compute costs as a significant issue. Specifically, as companies move beyond basic chatbots to adopt autonomous operational agents, costs are shifting from fixed SaaS expenses to variable consumption-based charges, eroding budget predictability. Most enterprises have already implemented or plan to roll out mitigation strategies—including token pooling, model downgrades, waste alerts, and restrictions on heavy users—to cut obvious waste rather than curtail AI adoption entirely. Several executives explicitly stated they are reluctant to restrict employees’ AI access, emphasizing their core goal is getting teams to actually use AI. To balance rising AI costs, these firms are opting to optimize other budgets by reducing external IT services, consolidating cloud spend, and similar measures. The report underscores that nearly all surveyed enterprises reported accelerating AI adoption rates, particularly among developer teams, confirming that cost increases are driven by usage growth rather than rising unit prices. UBS frames this dynamic as standard enterprise cost management, not a signal of stalled AI adoption. For example, Uber, which exhausted its full annual AI budget in a single quarter, still maintained generous token limits and scaled AI applications while offsetting costs by boosting engineer productivity. Looking ahead, UBS notes that AI model providers and hyperscale cloud service providers are ramping up efforts to improve token efficiency, which could cap recent price hikes and reshape market share among cloud vendors. Google Cloud and AWS hold an advantage here due to their in-house chips and custom models, the report adds. Additionally, enterprises’ growing discomfort with usage-based pricing models may put extra pressure on the non-AI software spending environment. Drawing on insights from a prior survey of 140 enterprise AI companies, the report finds that “unclear return on investment” remains the top barrier to AI adoption, while “lack of budget” has not cracked the top five. However, as token costs become more prominent, this cost dynamic is emerging as a key factor driving enterprises to pragmatically optimize their AI deployments.