Chinese AI models are now up to 90 per cent cheaper to run than leading US rivals, rattling Western AI economics.
Juniper Research found that Chinese AI models are rapidly changing where developers buy AI.
OpenRouter, an open marketplace where developers choose among competing AI models, recorded a sharp shift away from American labs. Google, OpenAI and Anthropic accounted for about 30 per cent of work conducted on the platform, down from about 70 per cent last year.
That poses an existential risk to Western AI outfits that have committed hundreds of billions of dollars to new data centres. Much of that money has been borrowed or raised through complex financing deals, based on customers paying premium prices for the best models.
Cheap Chinese alternatives shove a crowbar into that assumption. If the cost of running AI keeps falling, the revenue needed to pay for AI infrastructure and data centre build-outs starts looking a bit knackered.
Juniper Research found that Western frontier labs still have the edge where accuracy and complexity matter most. That includes long tasks, regulated industries and software running with little human supervision, but users pay a chunky premium for that reliability.
Two markets have emerged. One competes on price, while the other competes on quality. The risk is that Chinese models keep improving and push more workloads away from Western inference providers.
Juniper Research senior analyst Jawad Jahan said: “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 data centre build-out weakens, and, correspondingly, the financing structures resting on that revenue.”
