Chinese AI Models Flood US Market as Startups Flee Skyrocketing Costs

From DeepSeek to Qwen, Chinese AI models gain ground as US companies grapple with unsustainable AI bills — startups and Apple alike are making the switch.

SAN FRANCISCO — American AI models lead the world in capability, but their cost is driving a growing migration to Chinese alternatives.

Flo Crivello's San Francisco startup Lindy.ai, which builds AI assistants for email and calendar management, was burning through cash on Anthropic's top-tier models. "By far, our No. 1 expense was Anthropic — more than payroll," Crivello told NPR, referring to his two dozen employees. Last month, Lindy migrated 100% of its traffic to DeepSeek, the Chinese AI lab.

"It was just 10x cheaper," he said, saving the company millions of dollars. "It was a very, very simple business decision."

Lindy isn't alone. Crivello says every founder he knows in the AI space has either switched to Chinese models or is actively considering it. Even Uber CEO Dara Khosrowshahi admitted the company "blew through our AI budget in a quarter" on a recent podcast.

The shift isn't limited to startups. On Wednesday, Alibaba confirmed its Qwen AI model will be integrated into Apple Intelligence for users in China, sending Alibaba's U.S.-listed shares up 4%. Caltech spinout PrismML also released a compressed version of Qwen that runs entirely on an iPhone 15, squeezing 27 billion parameters from 54 GB down to under 4 GB.

U.S. lawmakers are taking notice, with Congress reportedly exploring curbs on Chinese AI model adoption by American firms. Meta was recently forced to dismantle a $2 billion acquisition of Chinese AI company Manus.

For now, economics is winning. With U.S. providers charging premium prices and Chinese labs offering models six to twelve months behind in capability but ten times cheaper, the calculus is shifting fast.

Sources: NPR, CNBC