UBS Survey: 60% of Enterprises Tighten AI Spending, Open-Source Models Like DeepSeek Seen as Beneficiaries
English summary
A UBS survey of about a dozen enterprise IT leaders reveals that roughly 60% of companies have implemented controls to tighten AI spending, driven by soaring token costs and insufficient ROI. The cost optimization push is causing some enterprises to significantly slow AI investment growth, though those in early AI adoption or with high returns remain less affected. Analysts expect model vendors like OpenAI and Anthropic to face the greatest near-term spending pressure, while open-source models such as DeepSeek and Chinese LLMs could benefit, especially for non-code use cases. Despite the pullback, UBS views the trend as a healthy industry adjustment, noting that enterprises are optimizing rather than halting AI use, and that next-generation models may further lower token costs.
Chinese summary
瑞银对十几位企业IT负责人的调研显示，约六成企业已出台措施收紧AI开支，词元调用成本激增和回报微薄是主因。成本优化使部分企业大幅放缓AI投入增速，但处于AI应用早期或能获得可观回报的企业受影响较小。分析师认为，OpenAI和Anthropic等模型厂商短期内将承受最大压力，而DeepSeek等开源模型及中国本土大模型有望受益，尤其在非代码类业务中。瑞银将此趋势定性为“良性调整阵痛”，企业并非叫停AI，而是转向高效利用，未来新一代模型还可能进一步压低成本。
Key points
About 60% of enterprises surveyed have imposed varying degrees of controls to tighten AI spending.
约六成受访企业已出台各类管控措施收紧AI开支。
Soaring token costs and insufficient ROI are the primary drivers of spending restraint.
词元调用成本飙升与AI投入回报微薄是企业收紧开支的主因。
Short-term pressure is expected on closed model providers like OpenAI and Anthropic; open-source models such as DeepSeek are likely to benefit.
OpenAI等闭源模型厂商短期承压，DeepSeek等开源模型有望成为受益方。
UBS characterizes the pullback as a healthy industry adjustment, not a halt, with cost optimization becoming a routine focus.
瑞银将此视为行业良性调整，企业优化而非停止AI投入，成本优化已成常态工作。