China’s AI startups leverage efficiency to undercut Western giants

While US firms grapple with the spiraling costs of training massive models, Chinese developers are transforming the artificial intelligence landscape through brutal optimization. Startups like DeepSeek and Moonshot AI are delivering performance parity with Western leaders while slashing operational expenses, turning a hardware scarcity into a competitive advantage.

European News·July 27, 2026·reads: 1,950

China’s AI startups leverage efficiency to undercut Western giants

The price gap is stark. Executing a standardized task on models like Kimi 2.6 or DeepSeek V4 Flash costs between two and 33 cents, compared to $2.75 for Anthropic’s Claude Fable 5. This shift mirrors China's historic path to industrial dominance, where the focus transitioned from sheer scale to extreme manufacturing efficiency.

US export restrictions on high-end AI chips acted as the primary catalyst for this pivot. Forced to operate with constrained hardware, Chinese engineers developed techniques to extract maximum utility from every unit of computing power. Selina Xu, China research lead for Eric Schmidt, notes that these startups have achieved a level of capital and computational efficiency that currently eludes OpenAI and Anthropic.

This aggressive pricing strategy has unsettled Western industry leaders, who warn that unregulated Chinese models could usher in significant security risks. Critics, however, suggest that these warnings are calculated maneuvers by OpenAI and Anthropic to stymie emerging competition as they prepare for public offerings.