Zte's xu ziyang discusses token valuation differences in AI billing for operators

At the 2026 China Unicom Partners Conference on September 10, ZTE President Xu Ziyang highlighted the challenges of handling traffic growth in the AI era. He noted that AI demands differ from previous technologies in three key ways. First, traffic is now largely driven by machines, with nearly 60% of internet traffic serving AI systems. Second, billing has shifted from bits to tokens, with token usage growing by a factor of 1,000 in 1-2 years. Xu emphasized that tokens from high-value models, like those used for poetry by Li Bai, differ in value from those generated by less skilled models, suggesting a need for token exchange rate differentiation in billing. Third, AI requires frequent context delivery, increasing uplink traffic and demanding deterministic networks for low latency. Xu called for further research into token exchange rates for accurate billing in the AI era. Xu also pointed out that AI’s edge intelligence collaboration requires reliable networks, pushing operators to adapt their infrastructure to meet these new demands.