China Mobile Reports AI Generative Traffic Surges Fivefold and Shifts Network Demands
China Mobile announced that mobile internet data traffic grew by 18% this year, with AI-generated content driving a fivefold increase and emerging as the main growth engine. Additionally, AI applications have shifted the network's uplink-to-downlink traffic ratio from the network-wide average of 1:9 to 1:6. The explosive growth in AI-driven traffic and shifting uplink-downlink ratios challenge traditional telecom infrastructure, which was primarily designed for download-heavy tasks. Understanding these changes is crucial for future network planning and reducing AI inference latency. China Mobile noted that current network resources struggle to match this structural data shift, and AI latency remains a bottleneck with the cloud platform side contributing 665ms compared to only 35ms on the mobile network side.
## BACKGROUND
Uplink flow refers to the data transmitted from a local device to the internet, whereas downlink flow represents data downloaded from the network. Time to First Token (TTFT) measures the waiting time from submitting a prompt until the model generates its very first response token, which is a critical metric for real-time AI user experience.