Chinese AI Labs Narrow Gap via Efficiency

Chinese AI labs are closing the performance gap with U.S. counterparts by optimizing algorithms for efficiency, not by acquiring more compute. U.S. agencies allege six Chinese firms, including DeepSeek and Moonshot, used bulk subscriptions to American AI outputs to train models, understating costs. Analysts say necessity from U.S. chip restrictions drove innovation in attention mechanisms, reducing computational complexity. Enterprise adoption is rising, with Chinese models handling 75% of engineering tasks at one-fifth the cost.