Nvidia backs Chinese AI models despite US regulatory concerns
In brief
- Nvidia optimizing support for DeepSeek V4 Flash, Qwen, and Kimi Chinese AI models
- Regulatory filing warns US restrictions could materially impact Nvidia's business
- Chinese AI models gaining global adoption through improved capabilities and lower costs
- Nvidia argues supporting diverse models keeps developers within US technology ecosystem
- Huawei and Alibaba competing by optimizing hardware for Chinese AI models
Hedging Against Regulatory Risk
Nvidia specifically cited DeepSeek, Qwen and Kimi as models for which restrictions on providing products and services could limit its ability to compete. The warning, filed in regulatory disclosures, reflects growing anxiety in Silicon Valley about Washington's appetite for AI export controls targeting China.
Chinese models have gained adoption among developers globally, including in the US, because of their improving capabilities and relatively low costs. That shift has created a dilemma for Nvidia. Abandon Chinese models and risk losing developers to competitors. Support them and invite regulatory scrutiny.
The Competitive Logic
Nvidia's defense rests on a straightforward argument: supporting models regardless of where they are developed gives users an incentive to remain within the American technology ecosystem. An Nvidia employee told CNBC that developers using both American and Chinese models are likely to choose the hardware and software stack for which those models are best optimized. If Nvidia doesn't optimize for Chinese models, developers may switch to alternatives.
This logic cuts both ways. The push also places Nvidia in competition with Chinese chipmakers including Huawei and Alibaba, which have been optimizing their own hardware for models such as DeepSeek and Qwen. Ceding the Chinese market entirely would hand those rivals a decisive advantage at home and potentially abroad.
Washington's Growing Pressure
Chinese open models have become a growing point of tension between Washington and Beijing. The concern isn't just about capabilities—it's about access. Open-weight models can be deployed anywhere, making export controls difficult to enforce.
Nvidia, Microsoft, Meta, Palantir and more than 20 other companies urged US policymakers in July to avoid premature restrictions on open-weight AI models. That collective push suggests the industry views blanket bans as economically damaging and technologically naive. But political winds can shift fast.
For now, Nvidia appears to be betting that optimizing for Chinese models while maintaining regulatory engagement offers the best path forward. Nvidia shares were up about 7.6% Thursday after the company reported stronger than expected second-quarter results and issued revenue guidance above Wall Street estimates, suggesting investors aren't penalizing the company's nuanced stance. Whether that continues if Washington tightens the screws remains an open question.
