The Zhitong Finance App learned that since the bank was first covered in August, the AI Token price index has dropped by about 25%, and the multi-regional average comprehensive price of OpenAI and Anthropic has dropped by about 31% and 47% respectively, which may redistribute the value of the AI industry chain. The bank anticipates that the AI Token price war will slow down model companies' annualized recurring revenue (ARR) growth in the short term. At this stage, companies with on-hand contract guarantees and infrastructure-as-a-service (IaaS) business exposure are preferred; in the long run, they are optimistic that the application and distribution process will be the beneficiaries of low-cost intelligence capabilities.
Yamato pointed out that the obvious drop in token prices mainly reflects an increase in the level of the low price model and the improvement in inference efficiency, but shifting the workload to a lower cost model will reduce the revenue achieved per token, and a significant increase in usage will be required to offset it. The bank believes that recent price cuts may offset a significant portion of the revenue increase brought about by short-term usage growth.
Furthermore, US AI vendors are gradually narrowing the pricing advantage of the Chinese model in the past and raising the model capability threshold. Currently, on average, the Chinese model is still competitive in price, but it no longer has a steady advantage in terms of cost performance.
In terms of stock selection, Yamato is optimistic about Alibaba-W (09988), Tencent Holdings (00700), Jinshanyun (03896), Smart Spectrum (02513), and MiniMax-W (00100), all of which gave a “buy” rating. The bank is optimistic about Alibaba's layout combining cloud business, self-developed chip and model business, and Tencent's broad distribution channels and cloud business development through proxy AI products.