【By Observer News, Liu Bai】 NVIDIA CEO Huang Renxun's statement that "China will win the AI competition" has once again sparked public attention on the competition between the US and China in this field.
"How many technologies in Silicon Valley are built upon China's AI foundation?" A Bloomberg article published on November 9 with this title cites multiple examples to illustrate that China's low-cost open-source AI models are not only attracting global users but also quietly gaining favor in Silicon Valley. The author believes that if the US government wants to maintain its lead for a long time, it should first think about why Silicon Valley is turning to China's AI.
Recently, Huang Renxun told the Financial Times: "China will win the AI competition." He then quickly added that China is only behind the US by "nanoseconds," and emphasized that the US must "accelerate forward and win over global developers" to maintain its lead.
The report states that although Huang Renxun's initial statement was overly simplistic, his pessimistic view of the US prospects is evident. Over the past year, he has consistently advocated that the US should allow his company to continue selling chips to the Chinese market, despite Washington continuously tightening export controls. However, he is worried that the "battle of developers" may have already gradually lost, and this concern is not baseless.
In recent weeks, a subtle shift has become increasingly evident. People had long speculated that China's low-cost open-source AI models might attract global users to switch from American products to Chinese ones. Now, signs show that this trend is subtly penetrating into Silicon Valley itself.
On September 24, visitors viewed Qwen at the 2025 CloudIn Conference. IC Photo
Investor Chamath Palihapitiya recently revealed in a podcast called "All-In" hosted by him and White House AI advisor David Sacks that one of his companies has moved most of its computing work to Kimi K2 model developed by Moonshot AI, a Chinese company. He said that this open-source model is "much cheaper than OpenAI and Anthropic."
Soon after, Brian Chesky, CEO of Airbnb, admitted that he did not connect his travel app with OpenAI's ChatGPT because the tools were "not yet mature enough." He said that Airbnb's new customer service system relies on more than ten different AI models, "with a large part relying on Alibaba's Qwen series: it's very good, fast, and cheap."
Given that Chesky has a close relationship with OpenAI CEO Sam Altman, his remarks are particularly noteworthy.
There are increasing cases of public recognition of Chinese AI models.
A startup founded by Mira Murati, former CTO of OpenAI, Thinking Machines Lab, stated in a blog post that its latest research "was inspired by the research results of Alibaba's Qwen 3 team and built upon it."
But what is more interesting is that this shift is taking root in a more hidden way.
Cursor, an AI programming leader valued at around $1 billion, launched a new version of its assistant last month. Afterward, there were numerous speculations online that it was built on a Chinese AI tool (such as DeepSeek). An investor even posted that when he used it, the assistant's internal monologue switched to Chinese.
Another US company, Cognition AI (also valued at around $1 billion), seems to have used Zhipu AI's foundational model Z.ai in its new programming agent. After social media users exposed this, Zhipu AI apparently confirmed it through a post, saying that this "highlighted the positive impact and value of open-source contributions to the ecosystem."
Neither of these two US companies responded to the email requests for comments from journalists. Due to the liberal open-source licensing agreements of Chinese AI models, companies can freely build products based on them.
Data compiled by the Hugging Face platform under the US open-source alliance "Atomic Initiative" confirms this trend: in terms of cumulative downloads by developers, Chinese models have surpassed American ones.
This change started slowly and then exploded rapidly: at the beginning of 2024, Meta's Llama had 10.6 million downloads, while Alibaba's Qwen had only 500,000; by last month, Qwen had reached 385.3 million, exceeding Llama's 346.2 million.
Additionally, systems derived from Qwen now account for more than 40% of newly released language models on Hugging Face, while Meta's share has dropped to 15%.
Regarding the phenomenon of international users flocking to Chinese AI tools, there have been geopolitical concerns, but for developers eager to launch products, especially in coding and software development, the so-called risks are no longer important compared to price and performance advantages. Moreover, open-source models can be downloaded, fine-tuned, and run locally, thus alleviating concerns about content and data privacy.
The author believes that it is still too early for Huang Renxun to declare the outcome of the AI competition. The US still has a clear advantage in cutting-edge chips and access to computing power, which are key elements for developing advanced systems.
However, China, driven by low-cost and open-source approaches, has indeed attracted more and more developers, who are the core force of AI innovation.
The article ends by reminding the US government: if they truly want to maintain their lead in the long-term competition, they should first ask: why is Silicon Valley starting to turn against?
Silicon Valley's "turning against" Chinese AI has had signs for a long time. In May this year, Huang Renxun specifically mentioned in a quarterly earnings call that Alibaba's Qwen model is the best among open-source AI models. Jack Dorsey, the founder of Twitter, also praised the Qwen series' code model Qwen3-Coder. Even Elon Musk, after seeing images generated based on Alibaba's Wan2.2, said the results were "indistinguishable from reality."
The New York Times published a long article on October 22 that directly stated: "Silicon Valley has fallen into fascination and envy of China."
The article wrote that a complex emotion of anxiety, envy, and self-reflection is sweeping through Silicon Valley, the center of American technology. Faced with China's rapid progress in infrastructure, artificial intelligence applications, and manufacturing at a "mechanical speed," many Silicon Valley elites are re-evaluating and admiring China's efficiency and execution capabilities, while reflecting on America's own problems in infrastructure, regulation, and manufacturing.
The article believes that American companies are racing to develop machines smarter than humans. However, if Silicon Valley delves deeper into China, they would find that China's AI industry is not obsessed with general AI. Chinese entrepreneurs focus more on applying artificial intelligence to services, devices, and manufacturing.
Former Google chairman Eric Schmidt and colleague Selina Xu have called on Silicon Valley to reduce its obsession with general AI, learn from Chinese counterparts, and integrate AI into daily life.
The author points out that Silicon Valley's envy of China has not only triggered discussions on work models and industrial policies, but also reflects the state of the United States itself, revealing the struggle of a country losing confidence.
"When China moves at full speed, moving goods, people, and information at a mechanical pace, we may be stuck in the past," said venture capital firm Andreessen Horowitz.
This article is an exclusive contribution from Observer News. Reproduction without permission is prohibited.
Original: https://www.toutiao.com/article/7570906733292356131/
Disclaimer: This article represents the views of the author and is not necessarily endorsed by Observer News. Please express your opinion below using the [Up/Down] buttons.