On 17 July 2026, Moonshot AI launched Kimi K3. A model with 2.8 trillion parameters. For the first time, a Chinese artificial intelligence surpassed American models on the benchmarks. Its score of 1,679 points on the Frontend Code Arena relegated Anthropic Fable 5 to 1,631 points and OpenAI GPT-5.6 Sol to 1,579 points. The same day, David Sacks, AI advisor at the White House, issued a warning: “This is how you lose the AI race.”
The shockwave is twofold. Technological, first. Economic, second.
Chinese models are not just powerful. They are up to 90% cheaper than their American competitors. The startup Lindy saved millions of dollars a month by switching from Claude to DeepSeek. DoorDash, the delivery app, made the same choice. The result: the share of tokens used on Chinese models has risen from 11% to over 30% on average, reaching 46% in some weeks, according to OpenRouter data relayed by China Daily.
American companies are voting with their wallets. They prefer Beijing’s open-source solutions to the closed, expensive models of Silicon Valley.
That is where the problem lies.
A few days after the launch of Kimi K3, the Trump administration announced it was considering restrictions. The Department of Commerce, the NSA and the White House Office of the Cyber Director are studying several options: adding Chinese models to the entity list, issuing security warnings, imposing supply chain rules. The objective is clear: to slow the rise of Chinese AI on American soil.
But this offensive is far from unanimous.
David Sacks, the White House AI advisor, has accused Big Tech of wanting to “eliminate open-source competition” through the government. Anthropic has proposed a statutory process allowing the government to block the deployment of frontier models that fail safety tests. A measure that, if adopted, could lock the market in favour of already dominant players.
“The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition,” Sacks stated, quoted by Axios. “The rest of the world won’t play by our rules if we bog ourselves down.”
The Trump administration is divided. On one side, the innovation advocates, led by Sacks, who want to let the market decide. On the other, the national security hawks, who are pushing for intervention. The measures under consideration range from a de facto ban to simple pressure campaigns.
Meanwhile, American companies continue to adopt Chinese models. And experts are asking questions.
Gary Marcus, an AI researcher quoted by the MIT Technology Review, believes that “Kimi K3 is largely on a par with the best American models and calls the business models of OpenAI and Anthropic into serious question.” In other words, it is not just a question of price. It is a question of business model. Chinese open-source solutions could render the closed, paid model of American Big Tech obsolete.
China, for its part, is observing with serenity. Lin Jian, spokesperson for the Foreign Ministry, stated that AI is “the common asset of humanity,” and that American restrictions reflect “anxiety over the erosion of US technological hegemony.” Chinese experts quoted by the Global Times believe that these measures could backfire against the United States by isolating it from global innovation.
The scenarios to watch.
A de facto ban, first. Chinese models are added to the entity list, blocking their access to American companies. A compromise, next. Targeted restrictions are imposed, but open-source models remain accessible. The status quo, finally. Internal divisions prevent any action, and the adoption of Chinese models continues to grow.
The United States faces a dilemma. Allowing Chinese models to thrive risks technological dependency. Banning them risks slowing innovation and strengthening American monopolies. The debate is open. The decision, however, is yet to come.
By the editorial team
Geostrat Watch – Deciphering the world to better anticipate.
Sources
https://big5.sputniknews.cn/20260720/1072404548.html
https://www.axios.com/2026/07/17/sacks-kimi-open-source-weights-trump
https://www.ixbt.com/news/2026/07/08/amerikanskie-kompanii-perekljuchajutsja-na-kitajskij-iskusstvennyj-intellekt–tak-deshevle.html
https://www.hindustantimes.com/world-news/us-news/trump-administration-weighs-blocking-us-access-to-chinese-ai-models-what-it-means-for-openai-anthropic-101784565266821.html
https://www.axios.com/2026/07/20/ai-us-china-open-source-kimi
https://www.newsnationnow.com/business/tech/ai/trump-administration-weighs-ban-chinese-ai/
https://www.technologyreview.com/2026/07/20/1140675/chinas-ai-models-have-trumps-ai-world-at-war-with-itself/
http://www.chinadaily.com.cn/a/202607/14/WS6a55aaaea310986e2b4652ce.html
https://www.worldjournal.com/wj/story/124277/9640602
https://peoplesdaily.pdnews.cn/business/er/30052716916
https://www.edgen.tech/zh/news/post/white-house-split-on-chinese-ai-as-28-trillion-parameter-model-emerges
https://www.tipranks.com/news/u-s-companies-seek-cheaper-chinese-ai-models-pressuring-microsoft-google-amazon-meta-and-nvidia-stocks
https://www.magzter.com/stories/dhnews/Hindustan+Times/Trump+administration+weighs+blocking+US+access+to+Chinese+AI+models+What+it+means+for+OpenAI+Anthropic-newsid-n720334109
https://udn.com/news/story/6811/9640122
https://www.chosun.com/english/industry-en/2026/07/21/J5JOQN64ZZFQ5P7OLZ4LN7A77M/