China's two biggest tech rivals, Tencent and Alibaba, are reportedly in talks for a joint investment in the country's top AI startup, DeepSeek.
This potential deal is significant because it signals a major shift from head-to-head competition to strategic collaboration. For years, these two giants have battled for dominance, so a joint investment is rare and suggests the AI landscape is forcing new alliances. It turns the race from 'who wins' to 'who can best use DeepSeek's technology' within their vast ecosystems—Tencent's WeChat and Alibaba's Cloud services.
So, why is this happening now? The causal chain points to three key factors. First is the critical issue of compute. Recent policy changes have reopened a path for Nvidia's powerful H200 AI chips to be supplied to China. The US and Chinese governments have given conditional approvals, reducing the uncertainty around securing the hardware needed to train and run large-scale AI models. This makes investing in a compute-hungry startup like DeepSeek a much safer bet.
Second is DeepSeek's own incredible momentum. In early 2025, its app briefly surpassed ChatGPT in US app stores, and Chinese businesses have been adopting its models at an 'unprecedented' scale. The company recently began talks to raise its first outside capital at a valuation around $10 billion, creating a clear opportunity for strategic investors like Tencent and Alibaba to get involved.
Third, both companies face immense strategic pressure. Tencent has already started integrating DeepSeek into WeChat search, proving its value. Alibaba, while advancing its own Qwen AI model, can't afford to be left behind if DeepSeek becomes the dominant model. Investing together is a smart hedge—it gives both access to top-tier technology and prevents a competitor from gaining an exclusive advantage. This move essentially neutralizes a threat while creating new opportunities for monetization through cloud services, ads, and AI agents.
- MaaS (Model-as-a-Service): A cloud computing service where customers can access and use AI models without building or maintaining the underlying infrastructure, similar to a subscription software.
- Compute: Refers to the computing power, primarily from specialized chips like GPUs (e.g., Nvidia's H200), required to train and operate large artificial intelligence models.