Is China quietly winning the AI ​​race?

Hundreds of millions of people flock to Pinterest every month for the latest styles.

One page titled “The Stupidest Things” is filled with crazy ideas to inspire creatives. I reused crocs as flowerpots. Cheeseburger-shaped eyeshadow. Gingerbread house made from vegetables.

But potential buyers may not know that the technology behind this product isn’t necessarily American-made. Pinterest is experimenting with Chinese AI models to sharpen its recommendation engine.

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

Of course, the San Francisco-based tastemaker could use any number of U.S. AI labs to get things moving behind the scenes.

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

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

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

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

Matt Madrigal, Pinterest’s chief technology officer, said the strength of these models is that companies like his can freely download and customize them, which is not the case with most of the models offered by U.S. rivals such as OpenAI, which makes ChatGPT.

“The open source technology we use to train our proprietary in-house models is 30% more accurate than leading off-the-shelf models,” Madrigal said.

And those improved recommendations can come at a much lower cost, in some cases 90% less than using the proprietary models favored by U.S. AI developers, he said.

Pinterest is not the only U.S. company relying on Chinese AI technology.

These models are gaining traction in many Fortune 500 companies.

Airbnb boss Brian Chesky told Bloomberg in October that the company was relying “heavily” on Alibaba’s Qwen to power its AI customer service agents.

He gave three simple reasons: “Very good,” “Fast,” and “Cheap.”

Further evidence can be found on Hugging Face, a place where people go to download ready-made AI models (including those from big developers Meta and Alibaba).

Jeff Boudier, who develops products on the platform, said cost is the reason young startups look to Chinese models over U.S. companies.

“If you look at the most popular models on Hugging Face, meaning the most downloaded and liked models by the community, Chinese models from Chinese institutes usually make up many of the top 10,” he told me.

“Some weeks, four out of five of Hugging Face’s top training models come from Chinese labs.”

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

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

But developers of Llama 4, which was released last year, are perplexed, and Meta is reportedly using open source models with Alibaba, Google, and OpenAI to train a new set of models for release this spring.

Airbnb also uses several models, including one based in the US, that it hosts securely on its own infrastructure. The company says the data is never provided to the developers of the AI ​​models it uses.

The consensus was that even as U.S. tech companies spend billions of dollars heading into 2025, Chinese companies threaten to take the lead.

“That’s not the case anymore,” Boudier said. “Right now, the best model is an open source model.”

A report released last month by Stanford University said that China’s AI models “appear to have caught up or surpassed” the world’s AI models, both in terms of their capabilities and the number of users.

Former British Deputy Prime Minister Sir Nick Clegg said in a recent interview with the BBC that he feels American companies are too focused on pursuing AI that may one day surpass human intelligence.

Last year, Sir Nick resigned from his position as head of global affairs at Llama developer Meta. Boss Mark Zuckerberg has poured billions of dollars into what he calls “superintelligence.”

Some experts now argue that these ambitions are vague and ill-defined, giving China an opening to dominate the open source AI space.

“Here’s the irony,” said Sir Nick. In the battle between China and the United States, “the world’s great dictatorships” and “the world’s largest democracies,” China is “further working to democratize competing technologies.”

The Stanford University report also suggested that China’s success in developing open source models may be partially explained by government support.

On the other side of the world, US companies like OpenAI are under intense pressure to increase revenue and increase profits, and are now turning to advertising to achieve that.

Last summer, the company released two open source models for the first time in years. But the company has poured most of its resources into its proprietary model to generate profits.

OpenAI boss Sam Altman said in October that the company is actively investing in securing more computing power and infrastructure deals with partners.

“Revenues are going to grow very quickly, but you should expect to invest heavily in training the next model, the next model, the next model, the next model,” he said.