Chinese hi-tech start-up DeepSeek recently shocked the world by releasing a new generative artificial intelligence (AI) model that, it claims, rivals US company OpenAI’s ChatGPT in performance while blowing it away in cost efficiency. The announcement was a boon to China’s confidence in its technological self-reliance and a blow to Washington’s conviction that it could contain China’s emergence as a superpower.
Talk of a “Sputnik moment” resounded in the US media. Yet the Cold War analogy missed a key point: Sputnik came during the early years of a protracted, high-stakes conflict that the world managed to survive. The timeline for an AI-powered Cold War 2.0 would not be nearly so forgiving.
Regardless of whether the fanfare about DeepSeek is true – and that’s still a big if – Washington must reconsider its failed, heavy-handed sanctioning of China and seek a more constructive approach to dealing with what is already the world’s second superpower. Put more simply, it’s time for the United States and China to talk.
Industry experts contest DeepSeek’s assertions that its AI model was built on relatively few lower-grade Nvidia chips for a mere US$6 million, in contrast to the hundreds of millions spent by Western companies. Testing has reportedly shown that the Chinese model may potentially be a copy of ChatGPT. Still, DeepSeek exists, and it is already changing how we use, deploy and apply AI models.
The notion that Washington could effectively curb China’s AI prowess was always wishful thinking. China has a decades-long track record of making cheaper technology. It has even done so quite recently. Now, it also can design such technology. As a result, China dominates the global electric vehicle, solar and robotics markets. And while China does not have Nvidia’s leading-edge chips, it does have an abundance of the rare earth minerals needed to produce them.
China has something else critical for AI success: size. Its 1.4 billion people, integrated into a finely tuned state surveillance system, supply all the data points its three largest AI players – Baidu, Alibaba and Tencent – need to train and run large language models. Those sources include an estimated 700 million cameras installed throughout the country, mobile apps linked to government servers and millions of daily financial transactions.
