DeepSeek disruption: How China’s AI moves wiped $500 billion off Nvidia’s market cap

Building on Matt Redley and Bob Huxford’s articles from Tuesday, David Linnett takes a deeper dive into DeepSeek and Nvidia. This week has been particularly eventful, with China’s announcement about DeepSeek causing a massive impact on US tech stocks, wiping over half a trillion dollars off Nvidia’s US market cap before the US market even opened.  

This announcement has clearly been designed to deliver maximum impact as it comes the week after the US announced an investment of $500 billion into AI. The amount is so large because training AI models is expensive, and they need large data centres to process the information. 

However, China has blind-sided the US and shown an alternative way to call on different parts of the AI to perform different tasks, rather than one AI doing everything.

OpenAI is fighting back and has claimed that China is plagiarising its models, which some people may consider as being extremely rich given that OpenAI is facing extensive claims of plagiarism and copyright infringement in training its own models. 

DeepSeek is potentially huge news as it smashes the narrative that AI will be driven by large US firms, which will be causing considerable alarm for US AI leaders, investors and chip suppliers such as Nvidia. 

There are obviously other factors to consider, such as privacy concerns and giving China access to US data. My advice at this stage would be to experiment with DeepSeek by all means, but be very careful about inputting any confidential information into the tool for now. 

And if DeepSeek’s language model wasn’t enough of a headache for the US, it has also announced the launch of the image generator JanusPro 7b, which it claims can outperform the best image generation tools offered by US firms.