[Soochow Securities Computer Wang Zijing] Recent Views on the AI Industry

English summary

1. On the AI bubble controversy. The current bull-bear divergence is essentially a timing difference between valuation expectations and industry realization, not a divergence on the long-term technological direction. The industry paths of China and the US differ significantly: overseas tracks see high capital expenditure coexisting with low investment returns, making local bubble characteristics more prominent; domestically, the orientation is toward implementation in the real economy, supported by cost-effective models and full-scenario applications, so the overall bubble risk is significantly lower than overseas. 2. On the policy direction for computing infrastructure. The 15th Five-Year Plan requires all newly built computing centers in various regions to fully adopt domestic chips, marking an upgrade of domestic computing autonomy from a guiding proportional requirement to a mandatory full-link coverage. On this basis, models such as 'computing-electricity coordination,' computing power going global, and the token economy continue to evolve, and independent controllability remains the core mainline running through the computing industry chain throughout the year. 3. On the path of frontier technology evolution. The AI technology stack is showing a full-level iteration trend, with disruptive technology directions emerging from interaction interfaces, software models, chip materials, to energy supply, physical site selection, and so on. ...