DeepSeek, a Chinese artificial intelligence startup, has found itself at the center of an unexpected whirlwind of attention after its breakthrough AI model, R1, captured the imagination of tech enthusiasts and industry heavyweights alike. The company, based in Hangzhou, China, had been gaining traction among engineers and AI enthusiasts for several days, as its innovative programming performance was compared to leading models like OpenAI’s ChatGPT and Microsoft’s offerings. At the World Economic Forum in Davos last week, DeepSeek’s name was mentioned among those pushing the boundaries of AI advancement, with some suggesting that its model was a key reason why nations like the United States were accelerating their own AI strategies.
DeepSeek’s sudden rise to prominence culminated in a remarkable surge in interest over the past 36 hours. Prominent Silicon Valley investors, including Marc Andreessen, and influential figures like Yann LeCun, the chief scientist at Meta Platforms, began to take notice. Andreessen praised DeepSeek’s AI model, calling it “one of the most amazing and impressive breakthroughs” he had ever seen. As the weekend progressed, the startup’s AI assistant rocketed to the top of Apple’s iPhone download charts and became one of the top downloaded apps on Google’s Play Store. However, this influx of users proved to be too much for the company’s systems, which led to a significant outage of more than an hour. The startup quickly responded by limiting sign-ups to users with mainland China phone numbers, citing “large-scale malicious attacks” on its services.
The attention surrounding DeepSeek has had profound ripple effects, particularly in the stock markets. The startup’s AI model, which it claims to have developed at a fraction of the cost of its rivals, contributed to a near $1 trillion drop in the stock market value of US and European technology companies. Notably, the share price of Nvidia, a leading AI chipmaker, plummeted by $589 billion, the largest single-day market value loss for the company. Some stocks, including Nvidia, saw a partial recovery during after-hours trading, but the fallout underscored the significance of DeepSeek’s disruptive entry into the AI space.
By Monday, the strain on DeepSeek’s systems was evident. On its status page, the company announced that it was limiting registration to users with mainland China mobile phone numbers. However, the startup did not clarify whether this restriction was temporary or how long it would remain in place. The outage was the longest in the company’s history and came as a result of unprecedented demand for its AI assistant.
Unlike many of its rivals, DeepSeek’s assistant offers users greater transparency by showing the reasoning behind its responses. This feature has been well-received by users, with positive reviews on both Apple’s App Store and Google Play Store. The company, founded by Liang Wenfeng, the chief of a quantitative fund, has developed an open-sourced AI model that is being hailed as a challenge to the vast sums of money companies have poured into AI research and development.
While it remains to be seen whether DeepSeek will continue to thrive as a cheaper alternative to more established AI models, the initial attention has raised important questions about the dominance of US tech giants in the AI space. Jun Rong Yeap, an analyst at IG Asia, noted that concerns are emerging over whether US tech companies’ pricing power is being threatened and whether their substantial AI investments require reevaluation.
However, like all AI models developed in China, DeepSeek is subject to self-censorship on politically sensitive topics, such as the Tiananmen Square incident, President Xi Jinping, or the possibility of China’s invasion of Taiwan. This self-censorship could be a point of contention for international users, who may find it jarring to interact with a chatbot that avoids such topics. Nevertheless, the initial success of DeepSeek serves as a counterpoint to the prevailing belief that cutting-edge AI technology requires massive amounts of computing power and energy—an assumption that has driven the soaring stock prices of companies like Nvidia and its suppliers.
While the exact costs of DeepSeek’s development and its energy consumption remain unclear, the company has presented figures suggesting that its model was developed at a fraction of the cost of its competitors. This achievement is particularly significant given the US’s increasing sanctions on Chinese companies, particularly those targeting Nvidia chips, which are essential for training AI models. Some experts believe that the rise of efficient AI models like DeepSeek could undermine the effectiveness of these sanctions.
Kai-Fu Lee, a renowned computer scientist, recently pointed out at the Asian Financial Forum in Hong Kong that China’s strength lies in its ability to engineer efficiently, especially when resources like computing power and funding are limited. According to Lee, the country has become adept at building AI models that are not only effective but also cost-efficient, challenging the assumptions of Western companies that have traditionally dominated the AI field.
Nvidia, for its part, has acknowledged the significance of DeepSeek’s breakthrough, describing its new model as an “excellent AI advancement.” The company also emphasized that DeepSeek’s model complies with US export restrictions on technology and illustrated how the Chinese startup has used a technique known as test-time scaling to create its AI model.
Despite the challenges it faces, including its controversial self-censorship and the recent technical difficulties, DeepSeek’s rise marks a turning point in the global AI race. Its efficient and cost-effective model poses a serious challenge to the established players, and its rapid growth highlights the increasing competition in the AI space, with companies like DeepSeek proving that innovation is not exclusive to the US or Europe. As the company continues to grow and refine its offerings, it could redefine the future of AI development and its role in shaping the global tech landscape.
