$Alibaba(BABA.US) Transferred from Information Equality: Software surged, no need for further explanation, open source and cost reduction make Agents and applications more deterministic, optimizing both revenue and cost, leading to higher ROI.

In the last trading session, Alibaba BABA was suddenly bought up from 1% to +6.7%. Foreigners are slow to react... Alibaba released Qwen2.5-Max (theoretical performance exceeds DeepSeek-V3) before the US stock market opened, and the stock price only began to respond at the end of the trading session.

Foreigners only noticed Deepseek and overlooked the equally excellent work of Kimi/Minimax/Alibaba Tongyi/Doubao. This time, Chinese AI is catching up as a whole, with the US stock market increasing by $10 trillion in market value due to AI. Is Chinese AI worth re-evaluation?

From a fundamental perspective, cloud vendors greatly benefit from this open-source victory. Model APIs are a cost item for cloud vendors, and closed-source clearly takes away a large portion of commercial interests, but open source can be wildly exploited by CSPs. North American CSPs can freely use DeepSeek, and even Meta can. Meta's previous goal of open-sourcing Llama was to allow the world to exploit it, but it aimed to siphon global development resources, build an ecosystem, and leverage global intelligence to optimize model directions. This time, however, Meta can immediately absorb DeepSeek's training methods for the next generation of Llama's training and optimization. This is the power of open source.

Of course, Alibaba is also in this position, and Tongyi's own model appears to be very powerful based on this release. One point that everyone overlooked is that a key prerequisite for DeepSeek r1's success is having an excellent solid base model like DS-V3, and once the subsequent RL training methods are established, they will be quickly referenced by other vendors. If Alibaba's Qwen-2.5-max indeed outperforms V3, there will be greater expectations for their RL inference model. A powerful model + Alibaba's computing power reserves + a complete cloud platform system, isn't that somewhat similar to the logic of buying North American CSPs last year?

If foreigners can react, domestic investors will certainly not lag behind. Just look at last night's Huatai DeepSeek conference call with over 2,000 participants on New Year's Eve... The hope is for a good start to the new year. Of course, there are also the earnings calls of several major CSPs on Thursday and Friday, but the marginal impact on Chinese AI has decreased; our own AI capital expenditure cycle is misaligned with North America, and from the beginning, the ROI has been healthier.

(Content has been shortened due to word limit)