DeepSeek's AI Revolution: Disrupting Global Tech Hierarchies

Chinese startup DeepSeek has launched AI models outperforming industry leaders at a lower cost, threatening to shift the technology landscape. Its DeepSeek-V3 model is reportedly trained with under $6 million in computing power, raising questions on U.S. AI investments and allegations of undisclosed resources.

In a seismic shift within the global technology landscape, Chinese startup DeepSeek has unveiled AI models that rival U.S. industry leaders but at a fraction of the expenses. The announcement has sparked widespread attention, notably after DeepSeek revealed that less than $6 million was needed to train its flagship model, DeepSeek-V3.

This development has catapulted DeepSeek's AI Assistant to the top of the U.S. Apple App Store rankings, surpassing OpenAI's renowned ChatGPT. The unexpected rise of DeepSeek raises questions about the hefty AI investments by major American tech firms, which have seen ramifications in stock markets, notably affecting shares of AI chip developer Nvidia.

Despite its success, the startup faces skepticism concerning its resource claims. Alexandr Wang, CEO of Scale AI, alleged that DeepSeek possesses 50,000 Nvidia H100 chips, information which purportedly violates U.S. export restrictions. DeepSeek has remained silent on these accusations, while financial analysts speculate its training costs far exceed stated figures.

(With inputs from agencies.)