UBS released a report on June 24 stating that Chinese AI companies are rapidly narrowing the performance gap with global leaders through targeted research and development, model architecture innovation, engineering optimization, and open-source collaboration, achieving significantly lower costs and positioning for increased global market share. The report attributes this advantage to cost efficiency becoming a critical variable as enterprise AI applications transition from experimental phases to scaled deployment. UBS Securities China Internet sector analyst Xiong Wei estimates Chinese model training costs at less than 10% of OpenAI and Anthropic's expenses, while average API pricing for major Chinese models remains below 20% of comparable global peers, yet Chinese frontier models continue closing the performance gap through rapid iteration while maintaining healthy profit margins.
Chinese AI Models Achieve Sub-10% Training Costs Versus Global Leaders
Xiong Wei stated that Chinese model training costs are estimated at less than 10% of those incurred by OpenAI and Anthropic, while average API pricing for major Chinese models sits below 20% of comparable global competitors. Despite significantly lower pricing, Chinese frontier models continue narrowing performance gaps through rapid iteration and intelligence level improvements, gradually increasing their share of global token usage while maintaining healthy gross margins. This indicates Chinese AI companies are not competing purely on low prices, but rather reshaping the cost curve of model services through engineering efficiency and scaled usage.
Global AI Market Expansion Creates Long-Term Share Gain Opportunity
Xiong Wei noted that the global AI market will expand rapidly, with long-term potential market space exceeding $10 trillion, providing substantial room for Chinese models to increase global market share. While enterprises previously prioritized the strongest models and maximum token usage when adopting AI, as AI applications enter actual commercial deployment, enterprises will place greater emphasis on return on investment (ROI), inference costs, latency performance, and sustainable usage scale. This shift will favor Chinese model suppliers with cost advantages and engineering implementation capabilities.
UBS Forecasts Stratified Model Market Favoring Cost-Efficient Providers
UBS expects the global model market to become increasingly stratified. Frontier models may still maintain pricing premiums for complex, high-value tasks, but in workloads with massive usage volumes and ROI sensitivity, models with stronger cost-performance ratios will be more widely adopted, creating long-term expansion opportunities for Chinese models. UBS draws parallels between the globalization potential of Chinese AI models and Chinese companies' overseas expansion experience in electric vehicles, smartphones, and home appliances; if Chinese models continue achieving breakthroughs in performance, pricing, open-source ecosystems, and localized deployment, the global AI competitive landscape may shift from dominance by a few high-priced frontier models toward a more diversified, stratified, and cost-sensitive market structure.
FAQ
What cost advantage do Chinese AI models have according to UBS?
UBS estimates Chinese model training costs at less than 10% of OpenAI and Anthropic's expenses, while average API pricing for major Chinese models remains below 20% of comparable global peers, according to analyst Xiong Wei's report released on June 24.
Why does UBS believe Chinese AI models will gain global market share?
UBS attributes the potential share gain to enterprise AI applications transitioning from experimental phases to scaled deployment, where cost efficiency and ROI become critical adoption variables, favoring Chinese models' cost-performance advantages in a global AI market with long-term potential exceeding $10 trillion.
