The United States surpasses China by 23 times in private AI investments, but the Kimi K3 model is changing the game.
According to the latest Stanford AI Index 2026 report, private investment in artificial intelligence in the United States reached $285.9 billion in 2025. For comparison, in China, this figure was only $12.4 billion — a gap of more than 23 times. However, against the backdrop of such financial dominance by the U.S., the Chinese model Kimi K3 from Moonshot AI has sparked a new wave of discussions in Washington about imposing strict restrictions.
Why is Kimi K3 causing concern in Washington?
Kimi K3 is distributed with open weights — anyone can download the model and run it on their own hardware. Its sudden success in programming tests caused a drop in shares of American chip manufacturers. Against this backdrop, officials have returned to discussing sanctions: last year, the U.S. Department of Commerce considered adding Chinese labs to the Entity List — the same trade blacklist that Huawei was placed on in 2019.
Moreover, the National Security Agency planned to publish a warning about threats from Chinese software, and the White House discussed the legal liability of American platforms for potential hacking of Chinese neural networks hosted on them. However, last time, supporting innovation was named a priority, and radical measures were abandoned. Now, hardline supporters have renewed pressure, and the Kimi K3 story has given them new arguments.
"This is a key moment for regulating artificial intelligence. Closed market leaders, who have already become a duopoly in revenue from AI models, want authorities to eliminate their open-source competitors," noted White House external AI advisor David Sacks.
Kimi K3's audience is growing rapidly: Moonshot suspended new subscriptions just 48 hours after launch, and the company is now preparing for an IPO in Hong Kong.
What does the funding gap hide?
The difference in investment volumes is impressive, but the report's authors acknowledge that official figures do not account for direct government injections from Beijing. From 2000 to 2023, Chinese government funds directed approximately $184 billion into the local AI sector. Financial benefits explain why businesses in the U.S. are increasingly choosing solutions from China. DeepSeek V4 Pro charges $0.87 per 1 million output tokens, while the same volume from Anthropic's Claude Fable 5 costs $50. The result is significant savings. For example, Coinbase CEO Brian Armstrong reported that the exchange's switch to GLM 5.2 and Kimi K2.7 Code halved corporate AI costs.
Washington has seen this before. In January 2025, the release of DeepSeek cost Nvidia a record $589 billion in market capitalization in a single day. Initial discussions of a potential ban emerged precisely after this crash. However, a ban may not work: Kimi K3's weights are already hosted in open repositories, and completely removing them is nearly impossible. Restrictions would only raise costs, and the launch of Alibaba's Qwen3.8-Max shows that new competitors will keep appearing regardless of bans.
Expert opinion: The financial gap between the U.S. and China in AI is enormous, but it does not guarantee technological superiority. The story of Kimi K3 and DeepSeek clearly demonstrates that the efficiency and accessibility of solutions can outweigh the volume of investments. The market is already voting with its dollars for Chinese models, and in the near future, we will see whether Washington can truly limit this flow or will prefer market mechanisms.