It’s more than DeepSeek

Significant VC activity and AI development opportunities are emerging in China. DeepSeek is the Vanguard of innovation from the artificial intelligence “moonshot” encouraged by the Chinese government. Not only will we see ongoing developments from Alibaba and Tencent, but there will also be a layer of elite AI companies at the forefront of China’s AI sector.

Fundamentally, Chinese companies are giving up on costly foundational model training to focus on building consumer-facing applications on top of other models. Research-focused firms continue to be driven in part by US semiconductor restrictions. Some AI foundational models include language, visual, video, and image processing and learning.

Research continues to develop efficient small language models, especially at Tsinghua University. DeepSeek is just the beginning. There is also a significant focus on developing the most advanced large language models (also at Tsinghua and private company offshoots). Substantial development has also occurred in AI infrastructure, especially models using combinations of different chip sets to coordinate AI tasks.

Heterogeneous computing clusters are developing as a challenge to US chip sanctions. Significant work is being done to increase the effectiveness of AI training in getting these heterogeneous chipsets to work in synchronization. Finally, they are working on overall efficiency for AI training and applications to reduce training time, capital expenditure, and technological reliance on the West.

Of course, what did we think would happen? The reaction has been increasing innovation and groundbreaking AI development activity in China. The sanctions and restrictions will amount to nothing and encourage accelerated advancement.