Chinese Firms Boost Nvidia H20 Orders Amid DeepSeek AI Boom

Chinese companies are ramping up orders for Nvidia’s H20 AI chip, driven by the rapid adoption of DeepSeek’s cost-efficient AI models.

This surge, reported exclusively by Reuters, highlights Nvidia’s continued dominance in the AI chip market, despite growing concerns over U.S. export restrictions.

Major Tech Firms Fuel Demand

Tencent, Alibaba, and ByteDance have significantly increased their H20 chip orders after DeepSeek’s AI technology gained global attention last month, according to sources.

These companies not only require chips for their own AI applications but also provide cloud computing services for other firms looking to integrate AI.

Smaller businesses in healthcare and education are also buying AI servers powered by DeepSeek models and Nvidia’s H20 chips, showing a broadening demand beyond traditional sectors like finance and telecoms.

U.S. Restrictions and Market Response

The surge comes amid potential new U.S. restrictions on the sale of the H20 chip to China. While such concerns may contribute to stockpiling, sources indicate that the main driver is DeepSeek’s AI revolution.

Despite initial fears that DeepSeek’s efficiency-focused AI models might reduce demand for high-powered chips, the opposite has happened.

The models’ ability to optimize inference-level computation has expanded AI’s integration into daily life, increasing the need for chips like Nvidia’s H20.

Nvidia Remains the Standard in China

While DeepSeek’s growth is expected to strengthen domestic chipmakers like Huawei, Nvidia’s H20 remains the industry standard in China. Analysts estimate Nvidia shipped around 1 million H20 chips in 2024, generating over $12 billion in revenue.

Growing AI Adoption Across Industries

Chinese companies are rapidly integrating DeepSeek’s AI into their platforms. Tencent plans to test the model within WeChat, and Great Wall Motors has incorporated it into connected vehicle systems.

As AI adoption continues to expand, demand for AI chips and computing power is expected to grow even further.