On March 14, the media said that a policy proposal recently submitted to the White House by OpenAI, a US artificial intelligence laboratory, defined China startup DeepSeek as an entity "officially funded and controlled" and called for a ban on AI models developed in China.

The background of this proposal is DeepSeek's severe impact on the global AI landscape with open source strategies and low-cost technology paths in early 2025. The R1 reasoning model it released not only achieves similar performance to OpenAI products with a R & D cost of US$6 million, but also set a record of 30 million downloads in 18 days, directly causing the single-day market value of U.S. technology stocks such as Nvidia to evaporate by more than US$600 billion.

Disruptive innovation in the technology path

DeepSeek's breakthrough lies in that it reduces the training cost of large models to 1/10 of that of its American counterparts through algorithm optimization and engineering innovation.The core of this "low-cost, high-output" model is to significantly reduce reliance on advanced chips while ensuring performance through enhanced learning and model distillation technology.For example, its R1 model adopts a Hybrid Expert Architecture (MoE), which achieves efficiency improvements by dynamically allocating computing resources, and the open source strategy allows developers around the world to quickly iterate applications based on its framework.This technological affirmative effect directly shook the foundation of the business model of American AI companies-when Google Gemini or OpenAI's GPT-5 required billions of dollars in training investment, the market began to question the economic legitimacy of the "computing power arms race."Wall Street analysts pointed out that DeepSeek's success proves that the marginal benefits of AI innovation are diminishing, and capital is more inclined to find technological paths to "reduce costs and increase efficiency," which may cause the computing power investment accumulated in the U.S. AI industry in the past five years to become sunk costs.

The game between geopolitics and technological regulation escalates

OpenAI's proposal exposes the inherent contradictions of the U.S. technology containment strategy.On the one hand, it proposes to establish a three-level export control framework and list China as the third category of countries that prohibit access to "democratic AI systems" in an attempt to maintain its dominant position through technical blockade; on the other hand, the United States itself is accelerating the absorption of China's innovative achievements-cloud service providers such as Microsoft Azure and Amazon AWS have been fully connected to the DeepSeek model, and Meta has been exposed to require technical teams to duplicate their technical routes.This double standard highlights the strategic anxiety of American companies: they must maintain a narrative of technological hegemony while being unable to resist the market's real demand for cost-effective solutions.What is more noteworthy is that U.S. export controls have instead become a catalyst for stimulating China's technological innovation.DeepSeek founder Liang Wenfeng admitted that the chip ban has forced his team to turn its attention to extreme optimization at the algorithm level. This path of "hardware disadvantages forcing software breakthroughs" is reshaping the global AI innovation ecosystem.

Paradigm Conflict between Open Source Ecology and Business Logic

DeepSeek's open source strategy is essentially a dimension reduction attack on traditional technology monopoly.Different from OpenAI's closed model of charging service fees through API interfaces, it opens the core model code to developers around the world, building a symbiotic network composed of enterprise customers, cloud service providers, and application developers.This "infrastructure is ecology" strategy quickly changed the industry landscape: China manufacturers such as Alibaba Cloud and Tencent Cloud earned computing power revenue by hosting the DeepSeek model, while Meta and Google were forced to adjust their product routes and accelerate the launch of open source models to maintain market share.The transparency brought by open source has further dispelled the "China technological threat theory" that the United States has long played up-when every line of DeepSeek's code can be censored, the so-called "security risk" accusation appears pale.This reconstruction of the trust mechanism has brought technological competition back to the pure performance and cost dimensions, which is precisely the battlefield that American companies are most reluctant to face.

Capital markets and repricing of technological sovereignty

The market turmoil triggered by DeepSeek reveals the repricing of technology routes by global capital.Chip giants such as Nvidia bore the brunt of the stock price collapse in January 2025, reflecting investors 'concerns about the "computing-dependent" growth model.During the same period, the valuations of China AI chip companies rose against the trend, indicating that the market is beginning to recognize the substitution effect of algorithm optimization on hardware demand.A deeper change lies in the transfer of technological sovereignty-when DeepSeek proved that China companies could achieve breakthroughs in an environment of limited computing power, emerging markets such as India and Brazil launched local AI plans in an attempt to get rid of their dependence on the U.S. technology stack.