MiniMax CEO Predicts 10x Drop in AI Inference Costs Within Two Years
Yan Junjie, CEO of Chinese AI startup MiniMax, predicted that inference costs for leading artificial intelligence models will decrease by another order of magnitude within the next one to two years. Speaking at the 2025 World Artificial Intelligence Conference in Shanghai, Yan attributed this projected efficiency gain to significant advances in system optimization and algorithms. This forecast follows a similar tenfold reduction in costs observed over the past year. His comments arrive amidst an intensifying price war in China’s large language model market, where custom project prices have plummeted from over RMB 10 million in 2023 to approximately RMB 200,000 today. Yan emphasized that AI capabilities are expanding while becoming more cost-efficient, driven by reinforcement learning in defined reward environments that accelerate generalization and expert-level performance. He also noted that the rise of open-source models and multi-agent systems is reshaping the industry landscape, making it unlikely for any single company to achieve total market dominance. MiniMax remains one of China’s six leading LLM startups, navigating this rapidly evolving and competitive technological sector.
Wire timeline
MiniMax CEO Predicts 10x Drop in AI Inference Costs Within Two Years
Yan Junjie, CEO of Chinese AI startup MiniMax, predicted that inference costs for leading artificial intelligence models will decrease by another order of magnitude within the next one to two years. Speaking at the 2025 World Artificial Intelligence Conference in Shanghai, Yan attributed this projected efficiency gain to significant advances in system optimization and algorithms. This forecast follows a similar tenfold reduction in costs observed over the past year. His comments arrive amidst an intensifying price war in China’s large language model market, where custom project prices have plummeted from over RMB 10 million in 2023 to approximately RMB 200,000 today. Yan emphasized that AI capabilities are expanding while becoming more cost-efficient, driven by reinforcement learning in defined reward environments that accelerate generalization and expert-level performance. He also noted that the rise of open-source models and multi-agent systems is reshaping the industry landscape, making it unlikely for any single company to achieve total market dominance. MiniMax remains one of China’s six leading LLM startups, navigating this rapidly evolving and competitive technological sector.
TechNode