Millions of people scroll through Pinterest every month in search of new fashion ideas, home décor inspiration and quirky design trends, often stumbling upon pages showcasing deliberately absurd creations such as Crocs turned into flower pots, cheeseburger-shaped makeup palettes or gingerbread houses built from vegetables. While the platform presents itself as a distinctly American tastemaker based in San Francisco, much of the artificial intelligence shaping what users see is no longer purely US-made. Pinterest has begun experimenting with Chinese-developed AI models to sharpen its recommendation and shopping tools, highlighting a broader shift under way in the global tech industry.
Pinterest chief executive Bill Ready said the platform has effectively evolved into an AI-powered shopping assistant, using machine learning to predict what users may want to buy or save next. Although the company could rely entirely on American AI firms, Chinese models have increasingly become part of its backend systems since the launch of DeepSeek’s R-1 model in January 2025. Ready described that moment as a turning point, noting that DeepSeek’s decision to make its technology open source triggered a surge of similar models worldwide.
Chinese players such as Alibaba’s Qwen, Moonshot’s Kimi and technology developed by TikTok owner ByteDance are now competing directly with US offerings. Pinterest’s chief technology officer Matt Madrigal said the appeal lies in the fact that these models can be freely downloaded, modified and tailored to a company’s specific needs. By contrast, most US-developed models, including those from OpenAI, remain largely proprietary. According to Madrigal, Pinterest’s in-house models trained using open-source techniques are significantly more accurate than many off-the-shelf alternatives and can cost up to 90% less to run.
Pinterest is not alone in adopting Chinese AI. Several major US corporations are increasingly relying on these tools, particularly where cost and flexibility matter. Airbnb chief executive Brian Chesky said last year that the company uses Alibaba’s Qwen extensively to power its AI-driven customer service features, citing its quality, speed and affordability. While Airbnb also employs US-based models, it hosts them within its own infrastructure and does not share customer data with the developers.
The growing popularity of Chinese models is also evident on Hugging Face, a leading platform for downloading and sharing AI systems. Jeff Boudier, who helps build products at Hugging Face, said cost considerations often push start-ups toward Chinese options. He noted that Chinese-developed models frequently dominate the site’s most downloaded and highly rated lists, with Alibaba’s Qwen overtaking Meta’s Llama models last year.
For years, Meta’s open-source Llama was considered the default choice for developers building custom AI applications. However, the release of Llama 4 reportedly disappointed many in the developer community, and Meta has since been working with a mix of open-source and proprietary models from multiple countries to train its next generation of systems.
As recently as early 2025, analysts widely believed Chinese firms were on the verge of overtaking US rivals despite America’s massive investment in AI. That narrative has now shifted. A recent Stanford University report found that Chinese AI models have largely caught up, and in some cases surpassed, their global competitors in both performance and adoption. Former UK deputy prime minister Sir Nick Clegg has argued that while US companies chase loosely defined goals such as “superintelligence,” China’s emphasis on open-source development is helping democratise the technology.
The Stanford report also pointed to strong government backing as a factor in China’s progress. Meanwhile, US firms such as OpenAI face mounting pressure to generate revenue, pushing them toward ads and closed systems. Although OpenAI released two open-source models last year, it continues to prioritise proprietary tools. Chief executive Sam Altman has said the company is investing heavily in computing power and infrastructure, betting that rapid revenue growth will fund future breakthroughs.
