Chinese artificial intelligence models now cost up to 90% less to run than leading US alternatives, according to Juniper Research, rapidly reshaping where developers purchase AI inference capacity.
The report, titled Beyond the Headlines: What the 'AI Bubble' Really Means, warns that this pricing disparity poses an existential risk to Western datacentre economics underpinned by hundreds of billions of dollars in committed capital expenditure.
Data from OpenRouter, an open marketplace where developers select between competing AI models, illustrates the scale of the shift. American laboratories including Google, OpenAI, and Anthropic accounted for approximately 30% of work conducted on the platform in 2026, down from around 70% just one year prior.
Independent reporting corroborates these figures, noting that Chinese-developed models captured all five top positions on OpenRouter by token volume in July 2026, led by Xiaomi's MiMo V2.5 and followed by offerings from DeepSeek, MiniMax, Alibaba's Qwen family, and Moonshot's Kimi.
"Open-weight models are closing the capability gap with frontier models at a fraction of the cost. Furthermore, they can run locally on consumer hardware. If this trend continues, the inference revenue underwriting the Western datacentre build-out weakens, and, correspondingly, the financing structures resting on that revenue," said Jawad Jahan, senior analyst at Juniper Research.
The analysis identifies a bifurcated market structure emerging: one segment competing on price and another on quality. Western frontier laboratories retain advantages in accuracy and complexity, particularly for long-context tasks, highly regulated industries, and software operating with minimal human supervision.
However, users pay a substantial premium for this reliability, with DeepSeek V4 Flash priced at approximately $0.14 per million input tokens compared to $5.00 for OpenAI's GPT-5.5.
This convergence creates structural financial risk. Western commitments to new datacentres assume customers will continue paying premium rates for top-tier models. If the price of running AI falls sufficiently, the revenue base supporting these infrastructure investments could erode.
By mid-2026, Chinese-origin models reached a weekly peak of 46 per cent of enterprise token volume on OpenRouter, with DeepSeek alone accounting for 17.6% of all routed tokens—more than any single US provider.
