Is China quietly winning the AI race?

Pinterest, the popular image-sharing platform, is increasingly turning to artificial intelligence technology developed in China to refine its recommendation engine. While many US-based companies are investing heavily in domestic AI solutions, Pinterest is leveraging the accessibility and cost-effectiveness of open-source models originating from Chinese tech firms.

The Rise of Chinese AI

Pinterest’s decision comes as Chinese AI models, such as DeepSeek R-1, Alibaba’s Qwen, and those from ByteDance, have gained prominence. The launch of DeepSeek R-1 in January 2025 spurred a wave of open-source AI development. Pinterest Chief Technology Officer Matt Madrigal noted that these open-source models can be freely downloaded and customized, a key advantage over proprietary models offered by US companies like OpenAI.

Did You Know? Pinterest’s recommendation engine influences the experience of hundreds of millions of monthly users, even guiding inspiration for projects like repurposing Crocs as flower pots or creating vegetable gingerbread houses.

Cost and Accuracy Advantages

The appeal of Chinese AI models extends beyond accessibility. Madrigal stated that using open-source techniques to train in-house models results in 30% greater accuracy compared to leading off-the-shelf models. Furthermore, the cost savings can be substantial, sometimes reaching 90% less than utilizing proprietary US-developed AI.

Broader Adoption Across US Companies

Pinterest is not alone in adopting Chinese AI. Airbnb CEO Brian Chesky revealed in October that his company relies “a lot” on Alibaba’s Qwen to power its AI customer service agent, citing its quality, speed, and low cost. The platform Hugging Face, a hub for AI models, shows a strong presence of Chinese models among its most downloaded and liked options. Jeff Boudier of Hugging Face noted that Chinese models frequently occupy four out of five of the top training model spots on the platform.

In September, Qwen surpassed Meta’s Llama as the most downloaded family of large language models on Hugging Face. Meta, which previously dominated the open-source space with its Llama models released in 2023, has reportedly begun incorporating open-source models from Alibaba and Google into its own development efforts.

Expert Insight: The increasing reliance on open-source AI, particularly from China, suggests a shift in the technological landscape. Companies are prioritizing practical benefits like cost and performance over exclusive, proprietary solutions, potentially reshaping the competitive dynamics of the AI industry.

A Changing Landscape

Experts suggest that the dynamic is shifting, with open-source models now considered the leading option. A Stanford University report published last month found that Chinese AI models have “caught up or even pulled ahead” of their global counterparts in both capability and usage. Former UK deputy prime minister Sir Nick Clegg observed an irony in this development, noting that China is “doing more to democratise the technology they’re competing over” despite being an “autocracy.”

While US companies like OpenAI continue to invest heavily in proprietary AI and explore revenue-generating avenues like advertising, the availability of powerful, cost-effective open-source alternatives from China presents a compelling alternative for businesses seeking to integrate AI into their operations.

Frequently Asked Questions

What is driving Pinterest’s decision to use Chinese AI models?

Pinterest is leveraging Chinese AI models due to their open-source nature, allowing for customization, and their significantly lower cost compared to proprietary models from US companies.

Which Chinese AI models are gaining traction in the US market?

DeepSeek R-1, Alibaba’s Qwen, and models from ByteDance are among the Chinese AI models gaining traction, with Qwen recently surpassing Meta’s Llama in downloads on Hugging Face.

How are US companies like Airbnb utilizing Chinese AI technology?

Airbnb CEO Brian Chesky stated his company relies “a lot” on Alibaba’s Qwen to power its AI customer service agent, citing its quality, speed, and cost-effectiveness.

As AI technology continues to evolve, will the balance of power in the industry shift further towards open-source solutions and international collaboration?