As the battle for AI supremacy intensifies, Trump White House reportedly contemplates a ban on Chinese models

By Jules Rimmer

Chinese artificial-intelligence lab Moonshot's launch of a cheap model poses a conundrum for the U.S. government

Moonshot's launch of its Kimi K3 model is focusing minds in the White House

The release of a cutting-edge AI model by the Chinese lab Moonshot has highlighted the high stakes at play in the battle for artificial-intelligence supremacy between the U.S. and China. A story published Monday on the Axios website reveals a secret battle being waged by the White House to fight off the threat of cheaper open-source models from its superpower competitor and protect its dominant players.

Axios's Maria Curi writes that the U.S. government is considering a ban on Chinese AI models. Citing "knowledgeable sources," the report indicates some parts of the Trump administration have tried to implement bans before with limited success, but the release of Moonshot's latest Kimi model has triggered a more urgent response.

Moonshot introduced Kimi K3 on Friday to coincide with China's World Artificial Intelligence Conference taking place through Monday in Shanghai. The launch of Kimi K3, which its developers claim offers similar performance to Anthropic's Fable and OpenAI's ChatGPT at a much lower price, has reignited concerns about cheaper Chinese technology undercutting the more expensive American versions.

Moreover, within the U.S., usage of more cost-effective Chinese open-source models is rising despite initiatives over the last year to discourage stateside adoption by identifying potential security threats, and the Axios report suggests bans have been considered.

There were disputes within the administration over whether such tight regulation would stifle innovation, according to the report. White House adviser Sriram Krishnan, who was opposed to a ban or government intervention, has left his position, which, according to the Axios report, suggests the hawks are now more prominent.

Rather than an explicit ban, however, a set of measures falling short of that, perhaps involving procurement rules, pressure campaigns and a push to highlight security concerns look likely to be pursued, according to Curi's reporting.

Curi quotes former AI and crypto czar David Sacks, characterized as holding influential opinions in Trump circles, who posted on X Sunday, "We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open source competition."

Sacks has warned on many occasions, according to Curi, against "regulatory capture" that would stifle competition and benefit the biggest labs. Perhaps understandably, those top AI developers are approaching the government every few months with the notion of banning open-source models.

A research note published Sunday by macro strategist Peter Tchir of Academy Securities considers the dilemma. Looking back on the process by which China since 2005 began to dominate global commerce, Tchir wonders if China is approaching the same moment in AI as it was then with industrial and consumer goods?

He concludes China is selling compute five to 10 times cheaper than the U.S., and, while its product may be inferior, it nonetheless boasts an obvious value proposition. When DeepSeek was launched, it suddenly impacted American AI companies profoundly because of the apparent cost benefits.

Tchir reports rumors that the Chinese models train themselves on U.S. models, reducing the time and cost involved, a process known as "distilling." Tchir isn't altogether sure this can be stopped.

What may also be extremely impactful for the U.S. labs is the possible reduction in their value, especially as the two largest, OpenAI and Anthropic, are planning to go public this year, with valuations of $1 trillion being bandied about for both. Moonshot itself is considering an IPO within the next six months, and the market valuation is allegedly more like $30 billion.

Moonshot's valuation has been rising rapidly, but its rumored IPO target of $30 billion would make it much cheaper than U.S counterparts.

Chinese models like Kimi K3, MiniMaxV2 pro and DeepSeek's V4 all offer services at a fraction of the prices charged by Anthropic and ChatGPT, and if their eventual stock-market flotations do, too, this could pose problems for the major U.S. players who need access to public markets to continue their ambitious expansion plans. OpenAI's most recent valuation was $852 billion, and Anthropic's $965 billion.

-Jules Rimmer

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

07-20-26 0944ET