[Zheshang AI Team Feng Cuiting] Why Is Capital Shifting to AI Mid- and Downstream? What to Buy in AI Mid- and Downstream?

English summary

Currently, there are four bearish logics for the AI upstream: 1. The logic of capital expenditure has reversed; the market no longer pays for giants' "long-term strategies" but instead focuses on return on investment, demanding actual improvements in cloud revenue and orders. 2. Debt risks are highlighted; tech giants rely on bond issuance for expansion, but the credit market is cooling. Amazon's new bond subscription multiple halved, and long-end tech bonds are under pressure. 3. Token Budgeting: the trend of cost reduction suppresses high-end demand; enterprises turn to low-price models, and the premium for cutting-edge high-price models shrinks. 4. The "shovel seller" logic is loosening; infrastructure demand such as chips and optical modules is bound to giants' spending cycles and cannot stand alone. We believe AI still has long-term value, but the market style is changing from "buying grand narratives" to "buying evidence of realization," along with narratives such as Token Budgeting benefiting open-source models and cloud vendors, as well as downstream AI application targets with token distribution logic. We highlight: 1. ...