Chinese AI Firms Undercut US Rivals With $2 Tokens as Asian Businesses Weigh Trade-Offs
Updated
Updated · CNA · Jun 15
Chinese AI Firms Undercut US Rivals With $2 Tokens as Asian Businesses Weigh Trade-Offs
2 articles · Updated · CNA · Jun 15
Summary
$2 to $3 per million output tokens from Chinese models such as MiniMax and Moonshot are drawing Asian businesses, versus about $9 for Google Gemini 3.5 Flash, $15 for Claude Sonnet 4.5 and $30 for GPT 5.5.
Token pricing matters more as companies shift from chatbots to AI agents that can require 50 to 100 internal operations per task, sharply raising usage costs for call centres, coding, research and back-office work.
A 50-person sales team could spend about $38,000 a year on GPT 5.5—roughly two to three times Chinese-model costs—while large enterprises using AI coding tools can face annual token bills approaching $1.5 million.
Experts said China's price edge comes from efficient architectures, cheaper energy, subsidies and open-source deployment options, but warned sticker prices can be misleading if models perform poorly in local languages or need more human review.
Asia is likely to become a multi-model market rather than pick one side, with US systems kept for premium reasoning and security-sensitive work while cheaper Chinese models handle high-volume routine tasks.
Will China’s low-cost AI strategy outmaneuver America's focus on premium models in the race for global dominance?
Are cheap AI tokens a breakthrough for businesses, or a hidden trap of poor quality and security risks?
The New AI Battleground: China’s Dominance in Token Pricing and Its Global Consequences
Overview
Since early 2026, the global AI market has shifted dramatically as Chinese firms aggressively leveraged token economics to gain a significant competitive edge. By strategically deploying cheap tokens, they created a substantial price gap with US models. This advantage is rooted in engineering choices like mixture-of-experts architectures, subsidized compute resources, and razor-thin margins. As a result, the battleground in AI is now defined by the cost and availability of AI tokens, not just model performance. This new landscape is rapidly changing how developers and businesses access and use AI worldwide.