Is China quietly winning the AI race?

Every month, hundreds of millions of users flock to Pinterest seeking inspiration for the latest trends, from fashion and home décor to culinary experiments. The platform’s curated pages, like one humorously titled "the most ridiculous things," showcase an array of imaginative ideas, from Crocs repurposed as charming flower pots and cheeseburger-shaped eyeshadow palettes to even a gingerbread house constructed entirely from vibrant vegetables. However, what these aspiring creators and potential buyers may not realize is that the sophisticated technology powering these personalized recommendations and visual discovery tools isn’t necessarily American in origin. Pinterest, a prominent Silicon Valley tastemaker, is actively experimenting with cutting-edge AI models developed in China to refine its recommendation engine, a strategic move that could signal a significant shift in the global AI landscape.

"We’ve effectively made Pinterest an AI-powered shopping assistant," the firm’s chief executive, Bill Ready, revealed in a recent interview. While Pinterest could undoubtedly leverage the vast array of AI capabilities offered by its domestic US labs, its recent embrace of Chinese AI underscores a growing trend. Since the groundbreaking launch of China’s DeepSeek R-1 model in January 2025, Chinese AI technology has become an increasingly integral part of Pinterest’s operations. Ready characterized this "DeepSeek moment" as a pivotal breakthrough, highlighting its significant impact on the open-source AI community. "They chose to open source it, and that sparked a wave of open source models," he explained.

This development places Chinese AI competitors such as Alibaba’s Qwen and Moonshot’s Kimi in a more prominent global position. TikTok’s parent company, ByteDance, is also actively developing similar advanced AI technologies. Pinterest’s Chief Technology Officer, Matt Madrigal, emphasized the substantial advantage offered by these Chinese models: their open-source nature. This allows companies like Pinterest to freely download and extensively customize them, a stark contrast to the proprietary models predominantly offered by US rivals like OpenAI, the creator of the widely recognized ChatGPT. Madrigal further elaborated on the tangible benefits, stating, "Open source techniques that we use to train our own in-house models are 30% more accurate than the leading off-the-shelf models." The cost-effectiveness of these open-source solutions is equally compelling, with Madrigal noting that they can reduce expenses by as much as ninety percent compared to utilizing the closed, proprietary models favored by many American AI developers.

‘Fast and Cheap’ AI Models Gain Traction

Pinterest is far from an isolated case; its reliance on Chinese AI technology highlights a broader trend of US enterprises increasingly integrating these advanced models. These Chinese AI systems are rapidly gaining traction across a diverse spectrum of Fortune 500 companies, demonstrating their growing influence and utility. Brian Chesky, the CEO of Airbnb, publicly acknowledged in October that his company heavily relies on Alibaba’s Qwen to power its AI-driven customer service agent. Chesky succinctly attributed this choice to three key advantages: the model’s exceptional performance, its speed, and its affordability.

Further compelling evidence of this trend can be observed on Hugging Face, a leading platform where developers congregate to download pre-built AI models, including those from major players like Meta and Alibaba. Jeff Boudier, who spearheads product development at Hugging Face, pointed to the cost factor as a primary driver for young startups opting for Chinese AI 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 – typically, Chinese models from Chinese labs occupy many of the top 10 spots," Boudier stated. He further noted the significant market presence, adding, "There are weeks where four out of five top training models on Hugging Face are from Chinese labs."

In a significant development last September, Alibaba’s Qwen model surpassed Meta’s Llama to become the most downloaded family of large language models on the Hugging Face platform. Meta had initially released its open-source Llama AI models in 2023, and prior to the emergence of DeepSeek and Alibaba’s offerings, they were considered the de facto choice for developers building custom AI applications. However, the subsequent release of Llama 4 last year reportedly left many developers underwhelmed. In response, Meta has reportedly been collaborating with companies like Alibaba, Google, and OpenAI to train a new suite of models, slated for release this spring. It’s important to note that Airbnb, while leveraging Chinese models, also utilizes several other models, including US-based ones, which are securely hosted within the company’s own infrastructure. According to Airbnb, the data used for training these models is never shared with the AI developers.

China’s Open-Source AI Success Story

Entering 2025, the prevailing sentiment within the tech industry was that despite substantial financial investments from US tech giants, Chinese companies were poised to take the lead in the AI race. However, this narrative has demonstrably shifted. "That’s not the story anymore," remarked Boudier. "Now, the best model is an open-source model." A comprehensive report published last month by Stanford University corroborates this assessment, concluding that Chinese AI models "seem to have caught up or even pulled ahead" of their global competitors, both in terms of their advanced capabilities and their widespread adoption.

In a recent interview with the BBC, former UK Deputy Prime Minister Sir Nick Clegg offered a critical perspective on the current state of AI development. He suggested that many US firms are excessively focused on the pursuit of artificial general intelligence (AGI) that could potentially surpass human intellect, a vision championed by figures like Meta’s CEO Mark Zuckerberg, who has committed billions of dollars to achieving what he terms "superintelligence." Some experts are now characterizing these ambitions as vague and ill-defined, creating an opportune window for China to establish dominance in the open-source AI sector.

"Here’s the irony," Sir Nick observed. "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’re competing over’." The Stanford report further posits that China’s remarkable success in developing advanced open-source AI models may be partly attributable to robust government support and strategic investment. Conversely, on the other side of the globe, US companies like OpenAI are facing intense pressure to demonstrate profitability and increase revenue, a challenge that has led them to explore new avenues, including advertising, to achieve their financial goals. OpenAI released two open-source models last summer, its first in several years, but has predominantly channeled its resources into developing proprietary models designed to generate revenue. OpenAI CEO Sam Altman conveyed in October that the company has made aggressive investments in securing extensive computing power and forging infrastructure deals with various partners. "Revenue will grow super fast, but you should expect us to invest a ton in training, in the next model and the next and the next and the next," Altman stated, underscoring a strategy heavily focused on future development and expansion, even at the expense of immediate profitability from open-source offerings.