中国AI悄然进军硅谷

China’s AI models are quickly gaining traction in Silicon Valley, becoming integral to the operations of American companies and earning the praise of a growing list of tech leaders.Their rapid ascent has highlighted the competitive edge that Chinese developers such as Alibaba, Z.ai, Moonshot, and MiniMax have been able to gain by offering so-called “open” language models at much lower costs than their rivals in the United States.

Recommended Stories list of 4 itemslist 1 of 4‘We have to fight’: BBC’s outgoing boss rallies staff amid Trump’s threatslist 2 of 4Ukrainian forces pull back under fierce Russian pressure in Zaporizhialist 3 of 4UK sentences Chinese scammer after record-breaking Bitcoin seizurelist 4 of 4Indigenous activists storm COP30 climate summit in Brazil, demanding actionend of listThe trend has also cast a critical glare on the US’s efforts to stunt China’s tech sector with export controls on advanced chips, which have not stopped Chinese developers from approaching the capabilities of Silicon Valley’s tech giants.

Airbnb CEO Brian Chesky generated headlines in October when he revealed that the short-term rental platform had opted for Alibaba’s Qwen over OpenAI’s ChatGPT, praising the Chinese model as “fast and cheap”.Social Capital CEO Chamath Palihapitiya revealed the same month that his company had migrated much of its work to Moonshot’s Kimi K2 as it was “way more performant” and “a ton cheaper” than models from OpenAI and Anthropic.

Programmers on social media also recently highlighted evidence that two popular US-developed coding assistants, Composer and Windsurf, were built on Chinese models.The assistants’ developers, Cursor and Cognition AI, have not publicly confirmed their use of Chinese technology and did not respond to requests for comment, though Z.

ai has said the speculation aligns with its “internal findings.”AI letters are shown on a laptop screen next to the logo of the Deepseek AI application in Frankfurt am Main, Germany, on April 1, 2025 [Kirill Kudryavtsev/AFP]Nathan Lambert, a machine learning researcher who founded the Atom Project, an initiative to promote open models in the US, said such public examples were the “tip of the iceberg”.

“Chinese open models have become a de facto standard among startups in the US,” Lambert told Al Jazeera.“I’ve personally heard of many other high-profile cases, where the most valued and hyped American AI startups are starting training models on the likes of Qwen, Kimi, GLM or DeepSeek,” Lambert said, adding that many US firms have been reluctant to publicly disclose their use of Chinese technology.

While it is not possible to precisely quantify the usage of different AI models, industry data points to the rising popularity of Chinese offerings.Chinese AI tools, including MiniMax’s M2, Z.ai’s GLM 4.6 and DeepSeek’s V3.2, took up seven spots among the 20 models with the most usage last week, according to data from OpenRouter, a platform that connects developers with AI models.

Among the top 10 models used for programming, four were developed by Chinese firms, according to OpenRouter.In the open model space, China’s clear lead is evident, with cumulative downloads surpassing 540 million as of October, according to an Atom Project analysis of data from hosting platform Hugging Face.

Rui Ma, the founder of Tech Buzz China, said Chinese models are particularly attractive to fledgling startups, while “high-resource organisations” have gravitated towards premium US models.“These are typically cost-conscious early-stage companies that experiment widely, and many of them will not survive,” Ma told Al Jazeera.

Unlike leading US platforms such as ChatGPT, China’s open-weight large language models make their trained parameters – called weights – publicly available.While open-weight models do not generate licensing or subscription fees, running them at enterprise scale requires large amounts of computing power, which creators can offer to users at a cost.

Developers such as Beijing-based Z.ai and Hangzhou-based DeepSeek have reported using older-generation chips that are not subject to US export controls, in relatively small quantities, dramatically reducing training and hardware costs compared with their Silicon Valley rivals.“The success of these Chinese models demonstrates the failure of export controls to limit China,” Toby Walsh, an expert in AI at the University of New South Wales, told Al Jazeera.

“Indeed, they’ve actually encouraged Chinese companies to be more resourceful and build better models that are smaller and are trained on and run on older generation hardware. Necessity is the mother of invention.”With lower input costs, Chinese firms have been able to offer their services far more cheaply than their US peers.

In an analysis published by AllianceBernstein in February, DeepSeek’s pricing for its models at the time was estimated to be up to 40 times cheaper than OpenAI’s, for instance.The logo of Chinese technology firm Alibaba is seen at its office in Beijing, China [File: Mark Schiefelbein/AP Photo]“I do think China’s AI progress has been underestimated, partly because the signal is fragmented,” Greg Slabaugh, a professor who studies AI at Queen Mary University of London, told Al Jazeera.

“Much of the uptake of Chinese models is in China. China’s scale in AI publications and patents has long been visible; the emergence of open-weight models simply makes that capability more globally consumable.”Some industry analysts have likened China’s approach to AI to the strategy undertaken by Chinese firms in other industries, such as solar panels, that flooded markets with cheap goods.

“This is the solar panel playbook running on software,” Poe Zhao, a Beijing-based tech analyst, wrote last week in his Substack newsletter, Hello China Tech.But while Chinese AI models have made inroads with their low cost, US tech giants are in a strong position to dominate the high-end market and highly regulated sectors where considerations such as national security are paramount, according to analysts.

Ma, the Tech Buzz China founder, said the development of AI could end up following a similar trajectory to the Android and iPhone platforms, the former of which has about three times as many users worldwide.“Over the longer term – likely faster than what we saw in the mobile era – it’s entirely possible that AI adoption might follow similar economic dynamics.

There are simply more users in the world who prioritise affordability than those who choose premium options,” Ma said.“But that doesn’t mean the greatest margins or market capitalisation will exist at the low end; value may still concentrate where differentiation, performance and trust command a premium.

”“In Fortune 500 and regulated sectors, widespread adoption is probably not imminent,” said Slabaugh, the Queen Mary University of London professor, referring to the uptake of Chinese models.“If there is a ‘rude awakening’, it may come on the pricing and flexibility front rather than from a sudden displacement of US models.

”