US Firms Pivot to Cheaper Chinese AI Models Amid Cost Pressures
A growing roster of US companies are adopting cheaper Chinese open-source AI models to manage rising token costs, challenging the pricing power of domestic leaders like OpenAI and Anthropic.
Major US companies are increasingly deploying AI models developed in China to cut operational costs, despite geopolitical tensions framing the technology as a bilateral race. DoorDash, Airbnb, and Siemens are among the firms experimenting with Chinese alternatives to power internal operations and consumer-facing tools.
The shift is driven by escalating expenses associated with US models. DoorDash co-founder and CTO Andy Fang noted this month that using a model from Chinese startup Moonshot AI offers "better quality" at a "cheaper cost." Fang announced on Wednesday that DoorDash is launching an experimental AI agent tool built with these economics in mind.
This is not an isolated corporate move. AI coding startup Cursor used Moonshot’s Kimi to build its Composer 2 agent, while Lindy reportedly replaced Anthropic’s tools entirely with DeepSeek’s V4 models. The trend is quantifiable: a March 2026 study by Hugging Face found that Chinese open-source models accounted for 41% of downloads on its platform.
For corporate buyers, the appeal extends beyond simple price tags to data control. Yasir Atalan, a data fellow at the Center for Strategic and International Studies, pointed to the flexibility of open-source architectures. "What we’re seeing right now is that it seems like the recent high-quality, high-performance models by U.S. companies seem expensive compared to Chinese models," he said.
Hosting these models locally allows enterprises to retain sensitive information rather than transmitting proprietary data to external providers. "It’s better for you to host a local model instead of just a closer model because that means everything will stay in that computer and will not go to any company," Atalan said. He noted, however, that this requires significant upfront capital, such as "$30,000 for GPUs, RAM, storage, etc."
The cost savings carry notable security tradeoffs. Snehal Antani, CEO of Horizon3.ai, warned that startups adopting these foreign systems "risk severe data sovereignty violations by exposing proprietary code and user data to foreign surveillance." He also highlighted potential "critical vulnerabilities in model integrity and reasoning."
Rather than a wholesale abandonment of US providers, the current phase appears to be one of tactical experimentation. Atalan suggested companies are segmenting their workloads, noting, "A company could try to use one of those open-source models for one task and use Claude for something else." For investors, this signals that US AI developers may face sustained pressure to lower prices or risk losing lower-stakes enterprise use cases to subsidized foreign competition.
