Is China quietly winning the AI ​​race?

Every month, hundreds of millions of users flock to Pinterest in search of the latest styles.

A page titled “the most ridiculous things” is full of crazy ideas to inspire creative people. Fangs converted into flower pots. Eye shadow in the shape of a cheeseburger. A gingerbread house made of vegetables.

But what potential buyers may not know is that the technology behind this product is not necessarily made in the United States. Pinterest is experimenting with Chinese AI models to improve its recommendation engine.

“We’ve effectively made Pinterest an AI-powered shopping assistant,” company boss Bill Ready told me.

Of course, the San Francisco-based trendsetter could use a number of US AI labs to run things behind the scenes.

But since the launch of China’s DeepSeek R-1 in January 2025, Chinese AI technology has increasingly become a part of Pinterest.

Ready calls the so-called “DeepSeek moment” a breakthrough.

“They chose to make it open source, and that started a wave of open source models,” he said.

Chinese competitors include Alibaba’s Qwen and Moonshot’s Kimi, while TikTok owner ByteDance is also working on similar technology.

Matt Madrigal, Pinterest’s chief technology officer, said the strength of these templates is that they can be freely downloaded and customized by companies like his – unlike the majority of templates offered by U.S. competitors like OpenAI, which makes ChatGPT.

“The open source techniques we use to train our own internal models are 30% more accurate than leading commercially available models,” Madrigal said.

And these improved recommendations come at a much lower cost, he said, sometimes ninety percent less than using the proprietary models favored by U.S. AI developers.

Pinterest isn’t the only American company relying on Chinese AI technology.

These models are gaining traction across a wide range of Fortune 500 companies.

Airbnb boss Brian Chesky told Bloomberg in October that his company relied “heavily” on Alibaba’s Qwen to power its AI customer service agent.

He gave three simple reasons: it’s “very good”, “fast” and “cheap”.

More evidence can be found on Hugging Face, the place where people go to download ready-made AI models, including those from leading developers Meta and Alibaba.

Jeff Boudier, who makes products on the platform, said it’s the cost factor that pushes young startups to look to Chinese models over their American counterparts.

“If you look at the top trending models on Hugging Face – the ones that are most downloaded and liked by the community – in general, Chinese models from Chinese labs take up many of the top 10 spots,” he told me.

“There are weeks when four of the top five training models on Hugging Face come from Chinese labs.”

In September, Qwen dominated Meta’s Llama to become the most downloaded family of large language models on the Hugging Face platform.

Meta released its open source Llama AI models in 2023. Until the release of DeepSeek and Alibaba models, they were considered the go-to choice for developers working on bespoke applications.

But the release of Llama 4 last year left developers disappointed, and Meta reportedly used open source models with Alibaba, Google, and OpenAI to train a new model expected to be released this spring.

Airbnb also uses several models, including US-based ones, and hosts them securely within the company’s own infrastructure. The data is never provided to the developers of the AI ​​models they use, according to the company.

Looking ahead to 2025, the consensus was that despite the billions of dollars spent by American technology companies, Chinese companies threatened to pull ahead.

“That’s not the story anymore,” Boudier said. “Now the best model is an open source model.”

A report released last month by Stanford University found that Chinese AI models “appear to have caught up with, or even outpaced” their global counterparts – both in terms of what they are capable of and the number of people using them.

In a recent interview with the BBC, former British Deputy Prime Minister Sir Nick Clegg said he believes American companies are too focused on finding AI that could one day surpass human intelligence.

Last year, Sir Nick left his role as head of global affairs at Meta, the developer of Llama. Boss Mark Zuckerberg has committed billions of dollars to achieve what he calls “superintelligence”.

Some experts now call these ambitions vague and ill-defined, giving China an opportunity to dominate the open source AI space.

“Here’s the irony,” said Sir Nick. In the battle between “the world’s great autocracy” and “the world’s greatest democracy” – China and America – China is “doing more to democratize the technology they compete for.”

The Stanford report also suggests that China’s success in developing open source models could be explained in part by government support.

On the other side of the world, American companies like OpenAI are under intense pressure to increase revenue and become profitable. and now turns to advertisements to help get there.

The company released two open source models last summer – its first in years. But he devoted most of his resources to proprietary models to help him make money.

OpenAI boss Sam Altman told me in October that he had invested aggressively to secure more computing power and infrastructure with partners.

“Revenues are going to grow very quickly, but the expectation is that we will invest heavily in training, in the next model and the next and the next and the next,” he said.