Last month DeepSeek, the Chinese artificial intelligence company, suddenly achieved world-wide fame with the launch of its open source R1 model, a significantly cheaper rival to the models available from Meta and OpenAI.
The impact was immediate with hundreds of millions of dollars wiped off the value of US tech stocks and Nvidia took a massive hit as a growing number of technology companies began to wonder whether they needed as many of its tools. This Chinese company appeared to have achieved comparable results to its rivals despite spending far less in terms of money and computing resources.
The massive investment being made into AI – and US tech companies had only recently pledged to spend billions of dollars more on investment into the sector – had been predicated on the idea that AI ‘hyperscalers’ would need huge amounts of computing power to train AI models. DeepSeek’s ‘success’ would suggest that this may no longer be the case.
For many this is good news. Advanced AI capabilities might now be achievable without the need for massive amounts of computational power, microchips and energy that had previously been thought.
Yet, within hours of its launch questions started to be asked of the model. OpenAI claimed that it had evidence that DeepSeek had used the American company’s AI model to train its chatbot, with Sam Altman suggesting that ‘distillation’ - which would violate OpenAI’s terms of service - had taken place.
Distillation is where one AI model asks repeated questions of another to train itself on how to respond.
Questions were also asked as to whether the claim that the company had spent just a few million dollars creating the RI was accurate, with some suggesting the true figure could be billions.
But probably of more significance is that a new front in the technological struggles between the US and China seems to have opened up as a result of DeepSeek’s AI breakthrough.
Commentators have argued that it shows that the US’s use of tariffs to restrain the technological rise of China is no longer viable. Chinese computer scientists have long had to take a more low-cost approach to development due to US trade restrictions on computer chips and DeepSeek’s announcement has made it clear that innovation can outmanoeuvre trade restrictions.
Tariffs may no longer be effective at curbing Chinese progress, yet while that strategy could be losing its edge the Trump administration is said to be looking at imposing further restrictions.
Whatever the decision the announcement by DeepSeek shows that China is now firmly establishing itself as a serious competitor to the US.
