US firms turn to Chinese platforms as costs surge

A growing number of companies in the United States are adopting Chinese AI models as the cost of using the most advanced US systems continues to rise.

Recent releases from Chinese developers, including DeepSeek and Z.ai, are being increasingly seen as competitors to frontier US models from Anthropic and OpenAI, narrowing the performance gap while offering significantly lower prices.

Released last month, Z.ai's GLM-5.2 ranked highly on several public benchmarks, trailing leading US models only narrowly while costing a fraction of comparable closed-source systems.

The trend is reflected in usage data from OpenRouter, a platform that provides developers access to multiple AI models. According to CNBC, citing OpenRouter figures, Chinese models have accounted for more than 30 percent of tokens used by US companies every week since Feb 8, peaking at 46 percent. That compares with an average of 11 percent over the previous 12 months and just 4.5 percent in the first half of last year.

Last month, San Francisco-based AI startup Lindy shifted all of its traffic from Anthropic's Claude models to DeepSeek.

"We did it, and you could see that cost curve go down, like, crash to the ground," Lindy CEO Flo Crivello told CNBC, adding that the move would save the company millions of dollars within months.

Ker Gibbs, former president of the American Chamber of Commerce in Shanghai, told China Daily that lower costs are an important factor, though falling unit costs are common in technology as production scales.

"The first semiconductor coming out of a fab (fabrication facility) was enormously expensive to make, but once the fixed costs were absorbed, every other chip was not expensive at all," said Gibbs, who teaches at the University of San Francisco. "What is new is the speed and scale at which the price differential has emerged in AI."

The cost advantage of Chinese AI models stems less from recovering fixed costs than from architectural efficiency, reflecting the same approach Chinese companies have used across other industries, he said.

"China's business people are pragmatic. They build products that are not just good — they are good enough. They don't overbuild or overdesign. Cost is always a factor," he said. "This is a familiar playbook applied to a new domain: achieve near-parity on performance, undercut dramatically on price, capture volume and let the market do the rest."

The Ramp AI Index, which tracks AI adoption by US businesses, found that the heaviest users spend about $7,500 per employee each month on the technology. It also identified DeepSeek as one of the fastest-growing AI vendors on its platform, suggesting companies are increasingly turning to lower-cost models.

Asked why businesses are paying more attention to AI costs, Gibbs said: "In short, the check arrived. The bills companies are paying for AI usage have come due and suddenly they realize what they have been paying.

"In some sense we're reaching the end of the beginning, which could be characterized as a bit of irrational exuberance and unrealistic expectations about the results AI would deliver," he said.

In the early stages of the AI boom, companies rushed to adopt the technology largely out of fear that competitors would gain an advantage, he said.

"The fear of getting left behind was stronger than any concerns about costs," Gibbs said. "Now that the bills are due, questions are being asked about the results and how to measure them."

As companies deploy AI to develop new products and improve efficiency, engineers are increasingly turning to lower-cost open-source and open-weight models, the most advanced of which now come from Chinese developers.

Unlike proprietary systems, open-source and open-weight models allow developers to inspect, use and, in some cases, modify key components of the software.