CHINESE AI NOW RUNS NEARLY HALF OF ALL ENTERPRISE API CALLS IN AMERICA — AND OPENAI AND ANTHROPIC ARE GETTING CRUSHED ON PRICE A CNBC investigation published July 7 put a number on something the AI industry had been quietly watching: Chinese-made AI models now account for between 30 and 46 percent of enterprise API token usage flowing through US developer platforms, according to traffic data from OpenRouter. For every single week since February 8, Chinese models have held at least 30 percent of the routed token share. At the weekly peak, they crossed 46 percent. The shift is almost entirely driven by cost. DeepSeek V4 Flash, a Chinese open-source model, runs at $0.14 per million input tokens. OpenAI’s GPT-5.5 costs $5.00, more than 35 times the price. The gap is not marginal; it is a different economic reality. DeepSeek alone now holds 17.6 percent of OpenRouter’s total token volume. Add Alibaba’s Qwen at 13.9 percent, and Chinese models collectively outpace all US-origin models on the platform. For US AI companies burning through capital to maintain a technical edge, this is an uncomfortable data point. The business case for paying a premium has always rested on superior performance. That case gets harder to make every quarter Chinese models close the gap. The security implications of routing enterprise workloads through foreign-origin models have not generated a meaningful congressional response. That silence is becoming harder to justify as the numbers compound. Keywords: Chinese AI models, DeepSeek enterprise usage, AI cost competition, US China AI race