The rise of Chinese AI company DeepSeek has sparked intense global debate in recent days. The model, which has topped the list of most downloaded apps on the US App Store - overtaking the previously dominant ChatGPT - has also caused the shares of several US tech companies to fall significantly.
On Monday, US President Donald Trump described the rise of DeepSeek as a "wake-up call" for the US industry. Meanwhile, U.S. congressional offices were informed that the use of DeepSeek was "unauthorized for official use in the House," Axios reported Thursday.
So why is the rise of DeepSeek causing such concern - even anxiety - in the US? A comment on social media from Yuyuan Tantian, an account associated with China Media Group, offers three main reasons based on the findings of the China Academy of Industrial Internet.
- Cost-effective disruption of the AI model monopoly in the US
DeepSeek redefines efficiency in AI development, delivering superior performance at a fraction of the cost. While OpenAI reportedly spent $78 million to train the GPT-4 model, DeepSeek achieved similar results for less than $6 million. With its pioneering approach to training, DeepSeek has lowered the barriers to entry in the AI market and made large-scale pre-training available beyond the tech giants.
In addition, the newly launched DeepSeek-R1 model is affordable - only $2.2 per million tokens - while the OpenAI o1 model costs $60 per million tokens. This breakthrough makes AI more accessible and opens up huge potential for research institutions, enterprises and knowledge-intensive industries.
With advances in both training and inference, DeepSeek is reshaping the AI landscape, setting new standards for efficiency and availability while challenging traditional models.
- Innovative approach sparks panic among US tech pros
DeepSeek has substantially reduced development costs with a new training approach. Unlike OpenAI, which relies on massive computational capacity, DeepSeek maximizes efficiency by using advanced algorithms to filter, summarize, and selectively process training data. This approach significantly reduces costs while improving performance.
The impacts are already being felt. Although Meta has put huge resources into training its Llama AI model, it has failed to outperform DeepSeek, which works at a fraction of the cost. This has prompted Meta executives to reflect on their spending and efficiency, sparking anxiety among U.S. tech professionals who fear for their jobs and expertise.
- Chinese AI models are gaining momentum
According to the China Academy of Industrial Internet, between Q4 2023 and Q1 2025, the gap in AI model capabilities between China and the world's leading companies has narrowed by nearly 75 percent. This suggests that the rise of DeepSeek is not an anomaly, but part of China's broader strategic progress in AI.
The report also highlights global trends in AI investment, with China receiving the second largest investment in AI at $5.5 billion - compared to $64.1 billion in the US. While this suggests that the US still leads the way in funding, it also highlights the significant scope for China's future growth in AI.
English 
Czech 
Spanish 
French 
Italian 
German 
Russian 
Chinese