The AI financial gap: The US outpaces China by 23 times in private investment, but the Kimi K3 model challenges Washington
According to a fresh report from the Stanford AI Index for 2026, the volume of private investment in artificial intelligence in the United States in 2025 reached $285.9 billion. For comparison, China attracted only $12.4 billion over the same period. The numbers seem to speak for themselves — the American AI market appears untouchable.
However, against the backdrop of this massive gap, the Chinese company Moonshot AI released the open-weight model Kimi K3, which instantly sparked a new wave of discussions in Washington about the need for strict restrictions. Within 48 hours of its launch, Moonshot was forced to suspend new subscriptions due to overload, and the company is now preparing for an IPO in Hong Kong.
Why Kimi K3 Became a Trigger for Regulators
The success of Kimi K3 in programming tests caused panic in the market — shares of American chip manufacturers plummeted. Against this backdrop, the U.S. administration once again began discussing adding Chinese labs to the Entity List — the same blacklist that Huawei was placed on in 2019.
Previously, alternative scenarios existed: 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 hosting Chinese neural networks. However, at that time, supporting innovation was prioritized, and radical measures were abandoned.
Now, proponents of a hardline stance have gained new arguments. White House AI Advisor David Sacks directly stated that closed market leaders, already profiting from AI models, want authorities to eliminate their open-source competitors.
What the Investment Gap Numbers Conceal
It is important to understand: official statistics 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. This fundamentally changes the picture.
But the key factor is not the volume of investment, but efficiency. DeepSeek V4 Pro charges $0.87 per 1 million output tokens, while a comparable volume from Anthropic's Claude Fable 5 costs $50. Coinbase CEO Brian Armstrong reported that the exchange's transition to GLM 5.2 and Kimi K2.7 Code halved corporate AI expenses.
Banning open weights is practically impossible — they are already hosted in public repositories. Any restrictions will only raise costs but will not stop distribution. And the launch of Alibaba Qwen3.8-Max shows that new competitors will keep emerging, regardless of bans.
Expert opinion: The market is entering a phase where the decisive factor is not the size of investments, but the ability to create efficient technologies at an affordable price. The U.S. maintains financial leadership, but Chinese open-source models are already changing the rules of the game, and regulation here is merely a palliative measure.