Emre Aytekin/Anadolu via Getty Images
A version of this article originally appeared in Quartz’s AI & Tech newsletter. Sign up here to get the latest AI & tech news, analysis and insights straight to your inbox.
The conventional wisdom about China's artificial intelligence sector has always been the same: good, but second place to Silicon Valley. Six months behind. Nine months behind.
That framing is now collapsing. The race itself has changed shape.
The clearest sign came in a two-week window in April, when three Chinese labs each released AI coding tools that matched the best Western models on standard benchmark tests, offered them as free, open software, and priced them below Western competitors.
The "months behind" gap, it turns out, depends heavily on which test you run and how you set it up.
This is not the story most people expected. A little over a year ago, a Chinese startup called DeepSeek sent shockwaves through global markets when it revealed its R1 model, built nearly as capable as the best American ones at a fraction of the cost. Nvidia $NVDA's stock briefly tumbled. Silicon Valley went into a quiet panic.
What's emerging now is something different, and in some ways more significant. A sprawling, fast-moving ecosystem where competition is pushing Chinese AI outward rather than just upward.
Open source as soft power
When DeepSeek released its latest model last month, offering discounts of 75% to 97% below comparable American models depending on the use case, the Western AI world mostly shrugged. No stock crash or hand-wringing this time. And that’s despite the prospect of high stakes IPOs for both Anthropic and OpenAI this year.
A year of headline-grabbing Chinese releases had raised the baseline for what counted as surprising. These tools have quietly become some of the most widely used openly available AI systems in the world. 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, using Chinese tech can run more than 90% cheaper than building on OpenAI products.
This was not the original strategy. Chinese companies initially turned to open-source AI because it was easier to build on foundations others had laid. But as the geopolitical stakes grew clearer, Beijing started to see openness as a feature, not a workaround.
Meanwhile, the U.S. moved in the opposite direction. As AI became extraordinarily expensive to build and treated more and more like a race, American labs increasingly shared less information. Even Meta $META, which had made open-source AI central to its identity through its Llama model family, began pivoting toward closed models after its Llama 4 release disappointed.
The irony cuts both ways. Google $GOOGL, Anthropic, and OpenAI have accused Chinese competitors of using a technique called distillation to extract capabilities from Western models without authorization, essentially learning from a teacher model to build a cheaper student version, an allegation Chinese officials say is groundless. Beijing, meanwhile, has shown it will protect its own assets too, blocking Meta's acquisition of the Singapore-based but Chinese-rooted AI startup Manus.
When adoption becomes the metric
Inside China, the story has shifted from building models to deploying them everywhere. More than 600 million people in China were using generative AI as of December, a 142% increase from the year before, according to the China Internet Network Information Center, a government-affiliated body.
The practical applications have spread in ways that are hard to compare directly to the U.S., where AI has been absorbed more quietly into existing tools and workflows. Judges in Shenzhen processed 50% more cases last year, partly with AI assistance. Tencent has embedded AI into WeChat, China's dominant messaging and payments app. Alibaba is offering AI that it describes as a “digital workforce” to merchants on its e-commerce platform.
But it's not all gas, no brakes. Last month, a Chinese court ruled that companies cannot fire workers simply to replace them with AI. The decision stands in contrast to the U.S., where policymakers have largely left workers to navigate AI displacement on their own. It was a signal that Beijing is trying to manage the social disruption it is simultaneously accelerating.
The remaining gaps are real, although it is mostly around semiconductors. U.S. export controls have kept the most advanced chips out of China, forcing labs to work around shortages in ways that add cost and time. Chinese chipmaker Huawei is growing fast, with AI chip revenues projected to hit $12 billion this year, but its hardware still lags the American frontier by at least two generations.
The picture that emerges is not of a country running a close second. It's of a country running a parallel race on a track it helped design. Chinese AI is cheaper, more open, more embedded in daily life, and closing fast.
Whether that constitutes being "behind" depends on which metric you value more, adoption or raw model capability.
