Chinese AI models achieve 90% of US performance while spending 82% less
This digest was compiled by AI from multiple sources — links to the originals are below.
Chinese AI companies have reached 90% of US model performance while spending only $124 billion compared to $694 billion by their American counterparts between 2023 and 2025, according to Jefferies. The 82% spending gap underscores the efficiency of firms like Alibaba, Baidu, Tencent, and ByteDance, which leverage architecture optimization to slash costs. This shift makes AI deployment increasingly affordable for smaller economies prioritizing rapid adoption.
The Spending and Performance Gap
US tech giants — AWS, Microsoft, Google, and Meta — invested $694 billion in AI from 2023 to 2025, while Chinese counterparts Alibaba, Baidu, Tencent, and ByteDance spent $124 billion, an 82% gap, according to Jefferies. Despite the disparity, top Chinese AI models now perform at approximately 90% of the level of their US counterparts. This includes MiniMax M2, which achieved 90% of GPT-5 Codex high and 106% of the open-source GPT-OSS-120B, demonstrating significantly higher performance per dollar.
Efficiency-Driven Cost Reduction
Chinese firms prioritize optimization of model architecture and compute efficiency over raw spending. DeepSeek cut API prices by 62% after efficiency gains in late 2025. This downward cost pressure makes AI deployment cheaper, a critical factor for companies and governments in smaller economies. As the cost of intelligence falls, the key resource becomes the speed at which organizations can integrate AI into productivity improvements.
What's Next
As AI model efficiency continues to improve, deployment costs are expected to drop further, intensifying competition between US and Chinese firms. It remains unclear whether the US can maintain its performance lead without matching the cost efficiency gains achieved by Chinese companies.
1 source
Chinese AI models achieve 90% of US performance while spending 82% less
