China’s daily token consumption hit ~500 trillion in June 2026, CAICT report says

The fourfold jump in daily token consumption over roughly six months points to demand scaling faster than the domestic serving stack has, which is why the TPS gap against foreign models matters more than the raw volume. That gap is the competitive pressure point for Chinese model providers and their infrastructure suppliers: throughput per second determines cost per token at this scale. The open question is whether domestic inference efficiency keeps closing on foreign benchmarks as consumption keeps compounding, or whether the performance shortfall becomes the binding constraint on how cheaply that demand can be served.