U.S. Startups Switch to Chinese AI Models to Cut Costs

Cost pressures drive AI model migration

Artificial intelligence spending has become the fastest-growing expense line for many U.S. businesses, and some are responding by abandoning premium American models in favor of Chinese alternatives that cost a fraction of the price.

Flo Crivello, founder of San Francisco-based Lindy.ai, told NPR that his company's Anthropic bill exceeded every other expense — including payroll for more than two dozen employees. Last month, Lindy migrated entirely to the Chinese model DeepSeek-V4, saving millions of dollars at one-tenth the cost.

"It was a very, very simple business decision," Crivello said, adding that every AI founder he knows is either considering the switch or has already made it.

The performance-cost tradeoff

While U.S. companies like Anthropic, OpenAI, and Google lead in AI capabilities, experts say Chinese models trail by six to 12 months. But China has established dominance in open-source models, which companies can download and adapt freely.

Eugene Cheah, CEO of San Francisco-based Featherless, compared the choice to selecting between a Ferrari and a Honda. "You can have the best luxury car, or you can just have a Honda at scale that works," he said.

For many applications — repetitive coding tasks, routine customer service, calendar management — the performance gap doesn't matter. Victor Su-Ortiz of Shanghai-based MiniMax said companies are moving away from "tokenmaxxing" (using as much AI as possible) toward cost optimization by routing different tasks to different model tiers.

OpenRouter, a platform providing access to multiple AI models, reported that DeepSeek usage jumped from 9% to nearly 20% since January. Models from MiniMax, Xiaomi, and Tencent have also gained traction.

Why it matters

The shift to Chinese AI models represents a fundamental challenge to U.S. companies' pricing power in a market they pioneered. As AI becomes essential infrastructure rather than a premium feature, cost will increasingly trump cutting-edge performance for most use cases. This dynamic could accelerate as major AI companies face pressure to demonstrate profitability ahead of potential public offerings — both Anthropic and OpenAI have filed confidential IPO paperwork.

Enterprise adoption extends beyond startups

The cost problem isn't limited to small companies. Uber CEO Dara Khosrowshahi said on a podcast that the company "blew through our AI budget in a quarter" for what was meant to cover the entire year. Bloomberg reported that Airbnb relied on Alibaba's Qwen model last year, which CEO Brian Chesky described as "good," "fast and cheap." Perplexity and Nvidia have also used Qwen.

Many companies avoid publicizing their use of Chinese models due to political sensitivities, but the models are widely available through platforms like Hugging Face and GitHub. Most users access them through U.S.-based hosting companies to keep data domestic.

The counterargument

Not everyone is ready to switch. Jon Gordner, CEO of newly founded Comment.io, said his company needs the best models available. "Saving a few dollars on a cheaper model isn't worth it if we have to spend two or three more weeks fixing its mistakes," he said.

Gordner noted that Anthropic and OpenAI are currently subsidizing users through discounted monthly subscriptions to build market share. But he acknowledged that once those subsidies end, "it's going to make a lot more sense to start evaluating Chinese models."

Ara Kharazian, lead economist at spending management company Ramp, predicted U.S. companies will respond competitively with their own cost-effective options. "The only reason why I'm bearish about the Chinese models is because I assume that the American model companies will respond competitively," he said.

These details were first reported by NPR.