China’s AI models are quickly gaining traction in Silicon Valley, becoming integral to how American companies operate and winning praise from a growing list of technology leaders.
Their rapid rise has highlighted how Chinese developers such as Alibaba, Z.ai, Moonshot and MiniMax can gain a competitive edge by offering so-called “open” language models at much lower costs than their U.S. rivals.
Recommended stories
list of 4 itemsend of list
The trend also casts a critical eye on U.S. efforts to stifle China’s high-tech industry with export controls on advanced chips, but have failed to prevent Chinese developers from accessing the capabilities of Silicon Valley’s tech giants.
Airbnb CEO Brian Chesky made headlines in October when he revealed that the short-term rental platform had chosen Alibaba’s Qwen over OpenAI’s ChatGPT, praising the Chinese model as “fast and cheap.”
Social Capital CEO Chamath Palihapitiya revealed in the same month that his company had migrated much of its work to Moonshot’s Kim K2 because it was “much better performing” and “much cheaper” than models from OpenAI and Anthropic.
Programmers recently highlighted evidence on social media that two popular coding assistants developed in the US, Composer and Windsurf, are built on Chinese models.
The assistant’s developers, Cursor AI and Cognition AI, have not publicly acknowledged the use of Chinese technology and did not respond to requests for comment, but Z.ai said the speculation was consistent with “internal research findings.”
Nathan Lambert, a machine learning researcher who founded the Atom project, an initiative to promote open models in the United States, said these public examples are “the tip of the iceberg.”
“China’s open model has become the de facto standard among American startups,” Lambert told Al Jazeera.
“I’ve personally heard of many other high-profile cases, where some of the most highly regarded and hyped American AI startups are starting to train models like Qwen, Kimi, GLM, DeepSeek, etc.,” Lambert said, adding that many U.S. companies are reluctant to publicly disclose their use of Chinese technology.
Although the usage of different AI models cannot be precisely quantified, industry data shows that Chinese products are becoming more popular.
Chinese AI tools, including MiniMax’s M2, Z.ai’s GLM 4.6, and DeepSeek’s V3.2, ranked seventh out of the 20 most used models last week, according to data from OpenRouter, a platform that connects developers with AI models.
According to OpenRouter, four of the top 10 models used for programming were developed by Chinese companies.
According to the Atom Project, which analyzed data from the hosting platform Hugging Face, China has a clear lead in the open model field, with more than 540 million downloads as of October.
Rui Ma, founder of Tech Buzz China, said the Chinese model is particularly attractive to fledgling startups, while “deep-pocketed organizations” are gravitating toward the premium U.S. model.
“These are typically early-stage companies that are cost-conscious and experiment widely, and many of them will not survive,” Ma told Al Jazeera.
Unlike major US platforms such as ChatGPT, China’s open-weight large-scale language models expose the trained parameters (called weights).
Openweight models are free to download and have no licensing fees, but they still consume large amounts of computing power when run at scale. Authors can offer that power to users for a fee through hosting services.
Developers such as Beijing-based Z.ai and Hangzhou-based DeepSeek use relatively small amounts of older-generation chips, which are not subject to U.S. export controls, and report significantly lower training and running costs than their Silicon Valley rivals.
Low input costs and computationally efficient models have allowed Chinese companies to offer hosting services much more cheaply than their U.S. counterparts.
An analysis published by AllianceBernstein in February estimated that DeepSeek’s model pricing at the time was up to 40 times cheaper than, say, OpenAI’s.
Toby Walsh, an AI expert at the University of New South Wales, told Al Jazeera: “The success of these Chinese models shows that export controls that restrict China are failing.”
“In fact, they have encouraged Chinese companies to be more resourceful and build better models that are smaller and trained and run on older generation hardware. Necessity is the mother of invention.”
Some industry analysts liken China’s approach to AI to the strategies of Chinese companies in other industries, such as solar panels, which have flooded the market with cheap products.
“This is a solar panel strategy that runs on software,” Beijing-based technology analyst Po Zhao wrote in Substack’s Hello China Tech newsletter last week.
“I think China’s AI progress has been underestimated, partly because the signal is fragmented,” Greg Slabaugh, an AI researcher at Queen Mary University of London, told Al Jazeera.
“Much of the adoption of the Chinese model is happening in China. China’s scale in AI publications and patents has been evident for some time. The advent of the open weight model has only made that capability more globally available.”
But while China’s AI models are gaining ground thanks to their low costs, U.S. tech giants are in a strong position to dominate high-end markets and highly regulated sectors where national security and other concerns are paramount, analysts say.
Tech Buzz China founder Ma said AI development could follow a similar trajectory to the Android and iPhone platforms, which have about three times as many users worldwide.
“In the long term, there is a good chance that AI adoption will follow similar economic trends, perhaps at a faster pace than the mobile era. There are simply more users around the world who prioritize affordability than those who choose premium options,” Ma said.
“But that doesn’t mean the highest margins or market capitalizations reside at the low end. Value may still be concentrated where differentiation, track record and reliability are valued.”
“Wide adoption is probably not imminent in Fortune 500 companies or in regulated sectors,” said Slabaugh, a professor at Queen Mary University of London, referring to the adoption of the Chinese model.
“If there is a ‘rude awakening’, it may be in terms of pricing and flexibility rather than sudden replacement of US models.”
