For the past two years, the prevailing narrative in the global artificial intelligence race has been one of American dominance. Driven by massive capital investment from Silicon Valley and the compute power of Nvidia, U.S. companies like OpenAI, Anthropic, and Google have consistently set the frontier of AI capabilities. China, constrained by export controls and a more cautious regulatory environment, was widely viewed as running six months to nine months behind.
A new analysis by Quartz argues that this framing is now collapsing and that “the race itself has changed shape.” The picture that emerges from the report, written by Jackie Snow, is not of a country running a close second. As Snow put it, “It’s of a country running a parallel race on a track it helped design.”
The Open-Source Pivot
The most striking evidence of China’s repositioning is in the open-source market. By the end of 2025, roughly a third of all global AI usage involved Chinese open-source models. For developers in Nigeria, Malaysia, and Brazil, building on Chinese technology can cost more than 90 percent less than building on OpenAI products. DeepSeek’s latest model, released last month, is priced at discounts of 75 to 97 percent below comparable American models.
In a two-week window in April 2026, three Chinese labs each released AI coding tools that matched the best Western models on benchmarks, and offered them as free, open software. This pattern of releasing competitive models at dramatically lower prices has become a defining feature of Chinese AI strategy, and it is reshaping the global developer ecosystem in ways that export controls on hardware cannot address.
The strategy has also created a powerful network effect. As more developers in the Global South adopt Chinese models, the feedback data flows back to Chinese labs, accelerating their improvement. The open-source approach is not merely a pricing strategy; it is a long-term play for global AI infrastructure dominance.
Scale of Adoption
The domestic adoption figures are equally striking. According to the China Internet Network Information Center, a government-affiliated body, 600 million people in China were using generative AI as of December 2025, a 142 percent increase from the year before. To put that in context, that is roughly double the entire population of the United States using AI tools on a regular basis.
The integration of AI into daily life in China is already producing measurable productivity gains. Judges in Shenzhen processed 50 percent more cases last year, partly with AI assistance. Tencent has embedded AI into WeChat, the messaging platform used by over a billion people. Alibaba is offering an AI “digital workforce” to merchants on its e-commerce platforms. As we reported earlier today, Baidu has proposed Daily Active Agents as the new defining metric for the AI era, reflecting a broader shift in how Chinese tech companies are measuring AI’s impact.
The Hardware Reality
The most significant vulnerability for China’s AI ambitions remains hardware. U.S. export controls have kept the most advanced chips out of China, and Huawei’s AI hardware still lags the American frontier by at least two generations, according to the Quartz analysis. This is not a trivial gap: training the largest frontier models requires the most advanced chips, and China’s inability to access them creates a ceiling on certain types of AI research.
However, even in this constrained environment, Huawei’s AI chip revenues are projected to hit $12 billion this year, a figure that would have been unthinkable three years ago. The domestic chip ecosystem is generating real revenue and real scale, even if it is not yet at the frontier.
The Quartz report also notes that Meta began pivoting toward closed models after its Llama 4 release disappointed, and that Beijing blocked Meta’s attempted acquisition of Manus, a move that underscores the degree to which AI has become a domain of active geopolitical competition rather than open collaboration.
The Distillation Debate
One of the most contentious issues in the U.S.-China AI relationship is the allegation of “distillation,” the practice of training a smaller model on the outputs of a larger one to extract its capabilities without authorization. Google, Anthropic, and OpenAI have accused Chinese competitors of using this technique to close the capability gap. Chinese officials have dismissed the allegation as groundless.
The debate matters because it goes to the heart of how the capability gap is being measured. If Chinese labs are able to rapidly approximate the performance of frontier Western models through distillation, then the six-to-twelve-month lead that U.S. officials cite may be more fragile than it appears.
As the AI race enters its next phase, the Quartz analysis suggests that Washington’s strategy of technological containment is yielding diminishing returns. Export controls can slow China’s access to the most advanced hardware, but they cannot prevent the diffusion of ideas, the development of algorithmic efficiencies, or the accumulation of the vast datasets that are increasingly defining the frontier of AI capability. The era of easy containment, the analysis implies, is over.