AI and the Unexpected Confrontation with the Crypto Assets Market

The trend of the integration of artificial intelligence and Crypto Assets is unfolding in an unexpected way. This dramatic development is not a synergy between the two, but rather the chain impact of AI on traditional capital markets and the Crypto Assets market.

Recently, China's AI model DeepSeek has emerged as a dark horse, with its download numbers surpassing ChatGPT for the first time, topping the US app store rankings. This breakthrough has drawn widespread attention from the global tech, investment, and media sectors, prompting people to consider the potential changes in the tech landscape between China and the US.

This event triggered a brief panic in the US capital markets. As a result, the stock prices of several tech giants, including Nvidia, ARM, Broadcom, and TSMC, saw significant declines. Even the Nasdaq 100 futures plummeted sharply, poised to record the largest single-day drop in recent times. It is estimated that the US stock market may have evaporated over $1 trillion in market value during the day's trading.

The Crypto Assets market closely follows the trend of the US stock market and has also seen significant declines. Bitcoin has fallen below $105,000, with a 24-hour decrease of 4.48%; Ethereum has dropped below $3,200, with a 24-hour decrease of 3.83%. Many investors are confused by this sudden crash, which may be attributed to lower expectations for interest rate cuts by the Federal Reserve or other macroeconomic factors.

The source of market panic may stem from DeepSeek's unique development model. Unlike companies such as OpenAI, Meta, or Google, which rely on substantial capital and extensive hardware resources, DeepSeek has achieved remarkable results in less than two years with just 200 employees and under $10 million in development costs. This efficient R&D model challenges traditional perceptions of AI development.

The success of DeepSeek is not only reflected in the cost advantages at the capital and technology levels but also challenges people's inherent perceptions. They have adopted a brand-new approach, significantly reducing the training costs of AI models by optimizing algorithms rather than simply increasing hardware investment. This innovative method has brought down the cost of training top AI models from $100 million to $5 million, reduced the required number of GPUs from 100,000 to 2,000, and decreased API costs by 95%.

This groundbreaking development overturns several traditional views, including perceptions of China's technological innovation capabilities, Silicon Valley's leading position in the AI field, OpenAI's technological advantages, and the massive investments required for AI model development. DeepSeek's success demonstrates that through clever engineering design, comparable or even better AI performance can be achieved with fewer resources.

Industry experts believe that the rise of DeepSeek represents an important victory for open-source models over closed-source models. This will promote the prosperous development of the entire open-source community and further enhance the AI technology level driven by open-source forces, including Meta. At the same time, it may also reduce enterprises' dependence on commercial APIs, providing greater development space for downstream applications.

However, experts also point out that the importance of computing power will not diminish for this reason. Historical experience shows that the enhancement of computing power is crucial for the development of AI. In the future, there may be more diversified reasoning chip products and a more prosperous application ecosystem for large language models.

Although the improvement in AI technology efficiency may reduce the resource demands of individual models, the overall market demand for computing power may actually increase. This phenomenon is similar to the Jevons Paradox during the Industrial Revolution, where improvements in technological efficiency led to an overall increase in resource consumption. As AI technology becomes more efficient and widespread, its range of applications will greatly expand, possibly leading to further growth in the overall market demand for computing power.