Open-source debut of China's DeepSeek: a signal to Trump, says AI expert

Goal could be to convince the US that semiconductor restrictions are fruitless

- Bhagyashree Garekar US Bureau Chief

WASHINGTON - The dramatic debut of Chinese artificial intelligence (AI) software DeepSeek, which triggered a deep sell-off in US stocks on Jan 27, was timed to land during US President Donald Trump's first week in office, according to an expert.

"That is not a coincidence," noted Dr Gregory Allen, director of the Wadhwani AI Centre at the Centre for Strategic and International Studies, speaking about the new open-source reasoning model which sliced the valuation of American chipmaker and stock market darling Nvidia by about US$600 billion (S$812 billion) in a matter of hours.

"DeepSeek's paper came out weeks ago, but they released the open version of the system during Mr Trump's first week in office," said Dr Allen, formerly the director of strategy and policy at the Pentagon's Joint Artificial Intelligence Centre.

"I am confident DeepSeek was told to release it during Mr Trump's first week," he said in an interview with The Straits Times.

DeepSeek released the open-source version of its DeepSeek-R1 model on Jan 20 - the very day Mr Trump was inaugurated. This release included a fully open-source model, a technical report and licensing for unrestricted commercial use.

The app rapidly became the most downloaded free app in the US market and 51 other countries. China accounts for 23 per cent of DeepSeek's total downloads, followed by the US at 15 per cent.

The DeepSeek AI model, which is said to at least match or even surpass OpenAI's ChatGPT on key benchmarks, operates at a fraction of the billions that US AI companies such as Meta, Alphabet and others have spent on Nvidia chips to upgrade their hardware.

Mr Trump said Chinese AI could act as a spur for the US companies to innovate.