Chinese AI models undercut US rivals by up to 90%, reshaping market share

September 2, 2026 at 06:10 PM2 min

This digest was compiled by AI from multiple sources — links to the originals are below.

Chinese AI models now cost up to 90% less to run than leading US alternatives, according to Juniper Research. Their combined share of OpenRouter workloads has fallen from around 70% to 30% for US firms like OpenAI, Google, and Anthropic. Juniper sees two distinct markets emerging: price-driven and quality-driven.

Key Facts

Chinese AI models can cost up to 90% less to run than leading US alternatives, according to Juniper Research.
US AI models from OpenAI, Google, and Anthropic saw their combined share of OpenRouter workloads fall from around 70% to 30%.
Juniper Research identifies two emerging markets in AI spend: a price-driven market and a quality-driven market.
Open-weight models are challenging the need to send inference to a cloud provider, as many small models can now be deployed locally.

Cost Advantage

Chinese AI models now cost as much as 90% less to run than leading US alternatives, according to a Juniper Research report. This cost gap is already affecting where developers choose to run their AI workloads. Juniper's research focuses on API platform and marketplace OpenRouter. Last year, around 70% of work on OpenRouter was conducted with US AI models from OpenAI, Google, and Anthropic. Their combined share has since fallen to around 30%, indicating a clear willingness to move workloads to cheaper alternatives.

Market Segmentation

Juniper says two distinct markets have started to emerge within AI spend. A price-driven market prioritizes cheaper inference, while a quality-driven market pays more for higher accuracy and reasoning. US companies still appear to dominate the quality-driven market. Emerging Chinese models deliver better on the price-driven front.

Open-Weight Shift

Open-weight models are challenging the idea that inference needs to be sent to a cloud provider. Many small models can now be deployed locally. Cloud providers could lose out on the major revenue they have experienced in recent years.