DeepSeek cut prices for two consecutive days. Based on their original compute scale, they definitely can’t support such a large user base, so it’s highly likely they’ve already raised funding and obtained compute support. There were earlier reports that they received money from Alibaba and Tencent, at a valuation of 200 billion USD.

If DeepSeek doesn’t raise more funding to secure compute support at this stage, they may really end up falling behind—and people will leave. Without compute, you can’t train top-tier models; with top-tier models, if you don’t have cards, you can’t do inference. Employee equity options only have value when the company has assets. After Luo Fuli left DeepSeek and went to Xiaomi, he was also able to push models to the front ranks, exceeding ByteDance and Tencent’s models—so the strength is terrifying.
Alibaba has chips ( Pingtouge ) and cloud infrastructure (Alibaba Cloud). In China, chip shipments rank second ( market share of 16% ); first is Huawei ( market share of 20% ). With the ability to connect domestically made chips for inference, Pingtouge can ensure DeepSeek’s future compute security.
Tencent’s AI products only had a chance to establish themselves to begin with by relying on a full integration with DeepSeek, and it also made open-source contributions for DeepSeek. Tencent’s own AI products aren’t very strong either—my guess is their compute hasn’t even been fully utilized.
ByteDance doesn’t invest in any external AI model teams; they build everything themselves. Although Huawei’s AI chips are ahead, the probability of investment is low. The biggest possibility is that DeepSeek has taken investments from both Tencent and Alibaba: one ensures what they can do now, and the other ensures what they can do in the future. Originally, the domestic large models Kimi, Minimax, and Zhipu were already invested in by Tencent and Alibaba; and now, with DeepSeek’s full family package, everything is complete.