<div id="readability-page-1"><div id="parent_top_div">By Catherine Thorbecke Nvidia Corp. Chief Executive Officer Jensen Huang recently declared to the Financial Times that: “China is going to win the AI race.” He later softened his stance, saying the US’s rival was merely “nanoseconds behind,” and that it’s vital America comes out on top “by racing ahead and winning developers worldwide.” Huang’s initial assessment is simplistic, and the motives for his negativity about the US are transparent. He has spent the past year arguing that America benefits from his company selling chips to the Chinese market despite Washington’s encroaching export controls. But he is right to worry that the battle for developers may already be slipping away. <div> In recent weeks, a subtle shift has become increasingly apparent. Speculation has been stirring for months that low-cost, open-source Chinese AI models could lure global users away from US offerings. But now it appears they are also quietly winning over Silicon Valley. ALSO READ: China tells technology companies to stop buying Nvidia's AI chips </div> Venture capitalist Chamath Palihapitiya recently said on his influential All-In podcast — co-hosted by White House AI czar David Sacks — that a company he works with has offloaded major workloads to Kimi K2, developed by Beijing-based Moonshot AI. The open-source model, he said, is “frankly just a ton cheaper than OpenAI and Anthropic.” Shortly after, Airbnb Inc. CEO Brian Chesky admitted that he didn’t integrate his travel app with OpenAI’s ChatGPT because the connective tools weren’t “quite ready.” Airbnb’s new service agent leans on more than a dozen different AI models, Chesky said. They are “relying a lot” on Alibaba Group Holding Ltd.’s Qwen lineup: “It’s very good. It’s also fast and cheap.” His comments are especially notable given Chesky’s close personal relationship with OpenAI CEO Sam Altman. The list of overt acknowledgements is growing. Thinking Machines Lab, the startup founded by OpenAI’s former Chief Technology Officer Mira Murati, said in a blogpost that its latest research was inspired by and built upon the work of Alibaba’s Qwen3 team. But what might be more interesting are the more subtle ways this shift appears to be taking root. Cursor, a much-hyped AI coding leader valued at some $10 billion, released a new version of its assistant last month. Internet conjecture has since mounted that it was built on top of a Chinese AI tool like DeepSeek, after a tech investor pointed out on X that it switched its inner monologue to Mandarin while he was using it. Another hot US-based company, Cognition AI Inc., also valued at around $10 billion, appears to have built its new coding agent off a base model from Zhipu AI, known internationally as Z.ai. After social media sleuths posted about their suspicions, the Beijing-based firm seemed to confirm their findings in a tweet, saying it “highlights the positive impact and value of open-source contributions to the ecosystem.” Both US companies didn’t immediately respond to my emailed requests for comment. But the permissive licensing nature of Chinese AI models means that they are free for companies to build products on top of. <div> Data from Hugging Face’s platform compiled by the ATOM Project, a US coalition in support of open-source AI, confirmed that. Chinese models have overtaken the US in terms of cumulative downloads by developers. The shift happened slowly, then all at once: In early 2024, Meta Platforms Inc.’s Llama had 10.6 million downloads to Alibaba Qwen’s meager half a million. By last month, Qwen had had 385.3 million cumulative downloads compared to Llama’s 346.2 million. And derivative systems built on Qwen now account for more than 40 per cent of new language models posted on Hugging Face, while Meta’s share has fallen to 15 per cent. </div> There have been geopolitical concerns about international users flocking to Chinese AI tools, specifically the potential for Communist Party ideologies to be embedded in outputs. But for developers racing to ship products, especially in the coding and software development sectors, these risks become less of a concern — especially when weighing for price and performance. Open models can also be downloaded, fine-tuned and run locally, mitigating content and data privacy anxieties. It’s premature for Huang to declare a winner. The US still has clear advantages when it comes to access to cutting-edge chips and computing power, critical components in developing advanced systems. But Beijing’s low-cost and open-source push is undoubtedly attracting developers, the backbone of AI innovation. If Washington truly wants to come out on top in the long run, it should start by asking why Silicon Valley is already switching sides. (Disclaimer: This is a Bloomberg Opinion piece, and these are the personal opinions of the writer. They do not reflect the views of www.business-standard.com or the Business Standard newspaper)</div></div>

英偉達

英偉達首席執行官黃仁勳對中國的人工智能進展表示擔憂，認為美國正在落後。最近的趨勢顯示，硅谷公司越來越多地採用低成本的中國人工智能模型，如 Kimi K2 和阿里巴巴的 Qwen，因其價格實惠且性能優越。像 Airbnb 首席執行官這樣的知名人物已承認使用這些模型而非美國同行。數據顯示，中國人工智能模型的下載量顯著上升，這引發了地緣政治的擔憂，但也突顯了開發者對成本和效率的關注。儘管美國在技術上仍然具有優勢，但向中國模型的轉變對美國在人工智能領域的主導地位構成了挑戰

How much of Silicon Valley's AI boom is powered by China's models?

- Nvidia CEO 黃仁勳表示中國將在 AI 競賽中獲勝，稱美國僅 “微秒” 落後。  
- 近期，中國低成本開源 AI 模型吸引全球用户，甚至贏得硅谷青睞。  
- 數據顯示，中國模型下載量已超越美國，開發者對價格和性能的關注加劇。  

硅谷的人工智能繁榮中有多少是由中國的模型推動的？