One trillion dollars gone. That’s what happened on January 27th, 2025. The most significant loss of market cap for a company in the history of the stock market. 600 billion dollars was shed for Nvidia, the American computer chip titan. But why? 

DeepSeek. The powerful new language model from China. A more competitive, efficient, and, more importantly, cheaper version of the prominently known American model OpenAI (ChatGPT). 

Arguably, the new Sputnik. DeepSeek was, to say the least, a smack in the face for American AI advancements. DeepSeek reportedly costs less than 6 million dollars to train. In comparison, it costs around 3-4 billion to run ChatGPT. But how did China achieve this impressive feat?

DeepSeek uses Nvidia H800 chips, less advanced than the most cutting-edge chips Nvidia offers. Ironically, the H100 and A100, the most ‘efficient’ chips, which are used for American language models, have been banned in China since 2022. Interestingly, DeepSeek uses an open-source framework. Researchers, developers–anyone–can improve, use, and modify the software. 

So, is this a problem? Yes and no. As an open-source, America and other countries can use DeepSeek as a reference for their language models. This is a good thing, as this collaborative environment allows companies to innovate the most efficient AI models. However, just a few days ago, Trump announced a 500-billion-dollar AI infrastructure plan with the likes of Oracle, Soft Bank, and ChatGPT, ensuring that the future of AI stays in the United States. DeepSeek has proven that’s simply not the case. Silicone Valley and the government must reconsider their place and the value of the AI market.