A report from EB SECURITIES states that Chinese open-source AI models have achieved a comprehensive lead in global Token usage. In terms of pricing, Chinese large language models hold a significant price advantage for Token output compared to their US counterparts. The power sector is currently at a cyclical valuation bottom. The catalyst of Token expansion overseas, underpinned by China's electricity price advantage, presents an opportunity for a sector-wide valuation recovery. The firm recommends investing in power operators that are involved in "computing-power" business and offer attractive valuations. The main points from EB SECURITIES are as follows:
Recent data from the online AI hosting platform OpenRouter shows that Chinese open-source artificial intelligence models, led by MiniMax, Moonshot AI, and Zhipu, have topped the global rankings for Token usage. Based on the latest data from this platform, Chinese open-source AI models have achieved a comprehensive overtake in global Token consumption.
The primary advantage of Chinese models lies in achieving performance comparable to top international models at an extremely low commercial cost. Current output prices for mainstream US AI models are all above $10 per million Tokens, whereas the output prices for China's mainstream AI models generally fall within the range of 10 to 20 RMB per million Tokens, representing a price advantage of nearly seven times compared to US models.
Driven by electricity price advantages and industry trends, Chinese large models possess strong competitiveness. Looking ahead, with the continued deepening of the "East Data West Computing" project, the accelerated formation of a national integrated computing power network, and the deep integration of clean energy with the intelligent computing industry, coupled with ongoing breakthroughs by domestic manufacturers in optimizing inference architecture and improving computing power scheduling efficiency, this competitiveness is expected to strengthen. From an electricity price perspective, China's overall electricity prices are highly competitive. In the first half of 2025, the green electricity settlement prices in Xinjiang, Gansu, Ningxia, Shaanxi, and Qinghai were 0.202, 0.265, 0.276, 0.312, and 0.228 RMB per kilowatt-hour, respectively. As the "East Data West Computing" project gradually materializes, it is expected not only to enhance the absorption of green power in western regions but also to potentially widen China's electricity price advantage for computing power. The firm believes this cost advantage is highly sustainable in the long term and will continue to be a core price competitiveness factor for Chinese models in the global market.
The development of "computing-power" business is steadily advancing, with multiple power operators accelerating their布局. Driven by the "East Data West Computing" project and a series of industrial policies, China has established a top-level design and implementation path for power-computing synergy, gradually improving a new energy supply architecture characterized by "direct supply as the main method, with the main grid as a backup," and promoting the coordinated layout of computing hubs and new energy bases. Against this backdrop, several power operators are actively engaging in "computing-power" synergy to explore a second growth curve: 1) Some are directly entering the data center field. For instance, Jinkai New Energy invested in and constructed the 5000P intelligent computing center in Changji, Xinjiang, and signed the Ulanqab intelligent computing center project. YUNENG CONTROLS plans to acquire a stake in a computing power company and take a controlling stake in an data entity. 2) Other operators are involved in the "computing-power" business by supplying power to data centers. Examples include Longyuan Power collaborating to deploy distributed computing nodes and regional industry computing networks, and Jinko Power Technology participating in a "wind-PV-storage" new energy microgrid demonstration project for a data center.
Investment recommendation: Suggested companies to watch include Fujian Funeng, Gansu Energy, Jinkai New Energy, YUNENG CONTROLS, and Fuling Electric Power. Risk analysis: Potential for slower-than-expected implementation of "computing-power" synergy policies and slower-than-expected advancement of AI development.