#DeepSeek冲击全球算力 The impact of DeepSeek on global computing power is mainly reflected in the following aspects:
Technological Innovation Level - Achieving High Performance with Low Computing Power: DeepSeek significantly reduces the computing power consumption of model training and inference through technologies such as structured sparse attention, mixture of experts systems, and dynamic computation routing. With only about 2000 chips and $6 million in costs, it achieves performance comparable to giants like OpenAI.
- Innovative Training Model: By adopting RL reinforcement learning strategies, AI learns to autonomously recognize the world, understand the laws of things, and conduct reasoning and exploration. Compared to traditional AI training methods, it can reduce ineffective training by 60%, and decrease parallel computing demands by 40%.
Market Structure Level - Changing Supply and Demand Relationships: With extremely low costs and excellent performance, it triggered a supply shock in the AI market, breaking the traditional reliance on high-performance computing resources and massive investments in the AI field, changing the market's perception of the high costs of AI computing power.
- Impacting the Status of Giants: Triggering a sharp decline in stock prices of tech giants like Nvidia on the US stock market, Nvidia's market value evaporated by nearly $600 billion in one day, causing a huge impact on technology giants that build computing power barriers.
- Lowering Industry Barriers: Its open-source model provides global developers with opportunities for collaborative innovation, breaking technological monopolies, allowing small and medium-sized companies and even individual developers to participate in AI competition, shifting market demand from 'customized by giants' to 'co-created by all'.
Industry Development Level - Reshaping Investment Logic: Investors' confidence in the profitability of the AI industry and the strong demand for high-tech chips has wavered, prompting a reflection on the rationality of massive investments and a reassessment of investment returns in the AI field.
- Indicating Development Direction: It proves that through methods like algorithm optimization, reliance on high-end computing power can be reduced, providing a new direction for the development of the global AI industry, prompting more companies to focus on algorithm innovation rather than purely pursuing computing power enhancement.
- Accelerating Industrial Transformation: With further technological development, AI is expected to break through application scenario bottlenecks, bringing about broader industrial implementation and promoting the acceleration of global AI industrial transformation.
International Competition Level - Impacting US Hegemony: The US had attempted to restrict China's AI development through export controls, but the emergence of DeepSeek suggests that such export restrictions are unlikely to stop the rapid development of Chinese AI technology, posing a potential threat to the AI hegemony constructed by the US.
- Enhancing China's Position: It may make China the de facto standard-setter for AI, allowing China to occupy a more advantageous position in future technology competitions, which is of significant importance in the Sino-US technological competition.
