DeepSeek’s R2 AI Model Is Reportedly Delayed After Chinese Authorities Encouraged the Firm to Use Huawei’s AI Chips; Beijing Is Still in Need of NVIDIA’s Alternatives

Aug 14, 2025 at 06:31am EDT
DeepSeek revolutionizing LLMs

Well, relying on Huawei's AI chips didn't go well for DeepSeek, as the AI firm has failed to train the R2 model on Chinese chips and is now opting for NVIDIA instead.

DeepSeek is Now Opting For NVIDIA's AI Chips To Train the R2 AI Model, Due to a Superior Tech Stack

The Chinese AI industry has relied heavily on NVIDIA and its AI ecosystem since the start of the hype. When the more recent US export controls came in, Beijing decided to speed up the development of domestic AI chip alternatives. This prompted the likes of Huawei and others to present their solutions to the Chinese Big Tech, such as the Ascend 910C AI chip; however, it seems like adopting homegrown alternatives isn't going too well.

Related Story NVIDIA RTX 6000D “Blackwell Pro For China” Tested: Features 17% Fewer Cores, 14% Less VRAM, & Lower Clocks

Financial Times reports that DeepSeek R2 is delayed due to using Chinese AI chips instead of NVIDIA's tech stack. The AI model was initially expected to launch in May. However, Huawei's AI chips struggle with stability issues and slower interconnect technologies. More importantly, there is no replacement for the CUDA tech stack for now, which means that eventually, firms like DeepSeek would need to rely on NVIDIA to train a superior AI model, and this is what is happening for now.

Huawei's AI chips are troubling Chinese firms right now, which means that the only other option they have is adopting NVIDIA's H20 AI chips, creating new demand for Team Green in the regional markets. However, apart from just demand, there is another issue for NVIDIA right now: the ongoing investigation into the presence of security backdoors in AI chips flowing into China. Beijing has reportedly advised domestic AI firms not to use Team Green's chips, as they contain mechanisms like location tracking.

It would be interesting to see how DeepSeek's R2 AI model, which is expected to debut in a few weeks, turns out. However, the report certainly shows that China's domestic alternatives are way behind in replacing NVIDIA, which means that for now, the local AI firms need to rely on chips from Team Green.

Follow Wccftech on Google to get more of our news coverage in your feeds.