Wire · regulatory
Chip shortages forced China to find way to make cheaper AI models than US
◆ Sectors
◆ Geography
◆ Source
◆ Verified
Fusion42 · 27 July 2026 · Fusion42 review
Chinese AI startups including DeepSeek, Moonshot AI, and Z.AI now offer models competitive with OpenAI and Anthropic but at 2-33 cents per task versus $2.75 for Claude, driven by chip export restrictions that forced efficiency innovations. US chip embargoes have catalysed a cost and performance advantage for Chinese models in inference workloads.
This Wire brief sits within Fusion42's coverage of AI Frontier Models, and 11 sources have reported it between 18 Jul 2026 and 5 Aug 2026.
◆ ◆ The Wire takeaway
Inference pricing has collapsed on constraint-driven efficiency. Your margin on selling access to Claude or GPT-4 to enterprises just disappeared—Chinese models run the same tasks for pennies, and that cost gap compounds as constraint-optimised architecture becomes the new standard, not the exception.
◆ Coverage
11 sources · first reported 18 Jul 2026 · latest 5 Aug 2026
◆ Related on Wire
◆ Topics