DoorDash and Airbnb are using models from China’s Moonshot AI, according to Coinpaper, which attributes the detail to an 11 October Washington Post investigation. The reported shift points to a practical challenge for US AI firms: competitive models can travel across borders even when chip exports are restricted.
Moonshot AI/Kimi Watch analysis
What happened
Coinpaper reports that American companies are adopting models from Moonshot AI and Alibaba, with Moonshot’s Kimi K3 attracting corporate users seeking lower-cost automation. The article also names Z.ai as pursuing a similar approach. It describes open-weight models as customisable and deployable without relying entirely on proprietary services.
The companies named are reported examples, not a measure of how widely Chinese models are being used across US business. Coinpaper’s account does not establish that every potential saving will survive the costs of computing, engineering and security.
Why it matters
The story separates two things Washington’s technology restrictions can easily blur: limiting access to advanced chips and limiting access to AI software. The reported use of Chinese models by US companies suggests chip controls have not stopped those models becoming part of business decisions.
For AI developers, cheaper alternatives may put pressure on prices. For companies choosing a model, the calculation is broader than the price of an API: customisation and deployment bring their own costs and responsibilities. A bargain model is only a bargain if it behaves well in the actual workflow.
Our read
This is a useful signal of competition on price, but not proof that US businesses are abandoning domestic AI providers. Coinpaper’s account points to specific adopters and models while leaving the scale of adoption unclear. The interesting question is whether cost and flexibility can outweigh the advantages of established providers, not whether a headline about “Chinese AI” settles the contest.
What to watch
- Whether more US companies disclose using Chinese-developed models.
- Whether the reported adopters share deployment costs or operational results.
- How US policy treats publicly available model weights alongside hardware exports.
Discussion spark: If a Chinese open-weight model is cheaper and can be customised, should a US company treat it as a serious option or prioritise domestic suppliers?
Sources and evidence
- Source update (11 October 2026, 16:52 UTC)
not affiliated with or endorsed by Moonshot AI and Kimi