Pinterest, DoorDash cut AI costs using Chinese open-source models

The growing use of Chinese open-source AI models by major US companies reflects a fundamental cost-benefit calculation that is now shaping the competitive landscape. Pinterest's disclosure that open-source models cost less than 8% of proprietary alternatives creates clear financial incentives for other enterprises to follow suit, especially as Wall Street has begun scrutinizing the capital expenditures of AI-heavy companies like Alphabet and Meta. The blending of models—American, Chinese, and others—into consumer-facing tools has outpaced regulatory clarity, leaving lawmakers concerned about intellectual property theft and potential bias while executives defend their fiduciary duty to shareholders.

The tension is sharpened by a disconnect between consumer sentiment and corporate behavior. Although Americans express strong distrust of Chinese AI (9% trust versus 52% for American models), most users cannot identify which model powers the chatbot they interact with when planning a trip or selecting a product. This opacity complicates both informed choice and regulatory enforcement. Congressional letters to Pinterest, DoorDash, Airbnb, and Cursor signal government intent to monitor the practice, yet executives report no measurable change in adoption rates. Luma AI's CEO observed that companies will likely maintain their current posture "until there are clearer guidelines," suggesting that without binding restrictions, cost savings will continue to outweigh geopolitical and trust concerns.