UBS recently held a summit with at least 19 AI companies to discuss the impact of Chinese open-source models and token cost optimization. These companies are leading players in their respective verticals and have significant influence on these issues. Following discussions with these companies, UBS drew several conclusions:
1. The AI industry is entering a "multi-model era." The future market will not be dominated by a few closed-source large models, but will instead form an ecosystem where cutting-edge models, open-source models, and enterprise-customized models coexist. The rise of Chinese open-source models is changing the competitive logic. AI competition is no longer just about pursuing the highest model performance, but has shifted to a comprehensive competition of "performance, cost, and efficiency."
2. Companies will dynamically choose between high-performance and low-cost models based on task complexity to reduce token consumption. This capability will become a core competitive advantage for AI application and infrastructure companies.
3. Industry-specific models based on fine-tuning of basic models will also become an important direction, driving AI from general-purpose capabilities to deeper applications in vertical scenarios.
4. Cloud vendors such as AWS and Azure will benefit from the multi-model ecosystem and the continuously growing demand for AI, with the impact being neutral to positive. 5. GPU demand will not weaken; AI workloads will expand to more inference tasks, which is generally beneficial to computing power providers such as Nvidia.
6. In contrast, the software industry may face pressure. To control AI costs, companies may reduce the use of some AI functions. At the same time, traditional software companies will find it more difficult to generate revenue through multi-model compatibility.
