Is China quietly winning the AI ​​race?

Although American companies have made great strides in AI, China’s focus on developing open-source AI models has made the country have a higher penetration rate among companies and startups and is close to winning the race.

Every month, hundreds of millions of users visit Pinterest to find the latest trends and styles. One page, titled “The Funniest Things,” is full of wacky ideas for inspiration: Crocs slippers turned into vases, eyeshadow shaped like cheeseburgers, or a gingerbread house made from vegetables instead of candy.

But what many users and potential buyers don’t realize is that the technology behind these offerings isn’t necessarily American. Pinterest is testing Chinese artificial intelligence models to improve its suggestion engine.

“We’ve practically turned Pinterest into an AI-powered shopping assistant,” says Pinterest CEO Bill Reddy. Of course, this brand can use any of the American artificial intelligence laboratories for behind-the-scenes processing.

But since the release of China’s DeepSeek R-1 model in January 2025, Chinese AI technologies have increasingly entered Pinterest. Reddy cites what he calls the “DeepSeek moment” as a turning point.

“They decided to open source their model, and that started a wave of open source models,” he says. Other Chinese competitors include Alibaba’s Qwen and Moonshot’s Kimi. Meanwhile, ByteDance, the owner of TikTok, is working on similar technologies.

Matt Madrigal, Pinterest’s chief technology officer, says the strength of these models is that companies like Pinterest can freely download and customize them; Which is not the case for most of the models of American competitors such as OpenAI, the creator of ChatGPT.

“The game text techniques we use to train our in-house models are 30 percent more accurate than the best off-the-shelf models,” says Madrigal. He adds that these improved offerings come at a much lower cost, sometimes up to 90 percent cheaper than using the proprietary models that American AI developers prefer.

Fast and cheap

Pinterest isn’t the only American company to rely on Chinese AI technology. These models are becoming increasingly popular among a wide variety of Fortune 500 companies.

Airbnb CEO Brian Chesky told Bloomberg in October that his company is relying “heavily” on Alibaba’s Qwen to power its customer service AI agent. He gave three simple reasons for this choice: this model is “very good”, “fast” and “cheap”.

More evidence can be found in Hugging Face; A platform where users can download ready-made AI models, including models from big developers like Meta and Alibaba.

Jeff Boudier, who works on product development at the platform, says the cost factor is driving young startups to Chinese models instead of American ones.

“If you look at Hugging Face trending models — the ones with the most downloads and popularity — usually Chinese models take up a lot of the top 10 spots,” he says.

“There are weeks when four of the top five training models in Hugging Face are owned by Chinese labs,” Boudier continues. In September, Qwen overtook Meta’s Llama to become the most downloaded family of large language models on the Hugging Face platform.

Meta released Llama open source models in 2023. Before the release of DeepSeek and Alibaba models, these models were the main choice of developers working on custom applications.

But last year’s release of Llama 4 didn’t impress developers, and Meta is said to be using open-source models from Alibaba, Google and OpenAI to train a new set of models to be released this spring.

Airbnb also uses several different models – including American ones – and keeps them securely on its own internal infrastructure. According to the company, the data is never shared with the developers of the AI ​​models.

China’s success in developing artificial intelligence

At the start of 2025, the consensus was that Chinese companies were on the verge of overtaking their Western rivals, despite billions of dollars in investment from US tech companies. “That’s not the story anymore,” Boudier says. “Now the best model is an open source model.”

A report published last month by Stanford University found that Chinese AI models “have caught up with or even surpassed their global competitors”; Both in terms of capabilities and the number of users.

Sir Nick Clegg, the former deputy prime minister of the United Kingdom, said in an interview that he thinks American companies are too focused on pursuing a form of artificial intelligence that may one day surpass human intelligence.

Last year, Nick Clegg stepped down as head of global affairs for Llama developer Meta. Meta CEO Mark Zuckerberg has put aside billions of dollars to achieve what he calls “superintelligence.”

Some experts now consider these ambitions vague and unclear; An issue that can give China the opportunity to dominate the space of open source artificial intelligence.

“Here’s the irony,” says Nick Clegg. According to him, in the battle between “the world’s largest autocratic government” and “the world’s largest democracy” – that is, China and the United States – China is “more than the United States democratizing the technology they are competing against.”

The Stanford report also says that part of China’s success in developing open source models can be explained 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 are now turning to advertising to achieve this goal.

The company released two open-source models last summer, the first in years. But he spent most of his resources on exclusive models in order to generate income from them.

OpenAI CEO Sam Altman said in October that the company has invested heavily in securing more computing power and infrastructure contracts with partners.

“Revenue will grow very quickly, but you should expect us to invest heavily in training and next-generation models,” he said.