Institutional forecast: China's hardware expenditure on generative AI related networks will increase from 6.5 billion yuan in 2023 to 33 billion yuan in 2028 IDC statistics show that in the second half of 2024, the daily average consumption of tokens for commercial implementation of large models in China increased nearly tenfold. In 2024, the cumulative call volume of public cloud large models in China reached 114.2 trillion tokens (excluding calls using overseas MaaS platforms). With the continuous increase in token usage in data centers, the significant growth in east-west traffic, and the continuous increase in communication energy consumption, the network is becoming a decisive bottleneck for AI infrastructure. In 2024, the shipment volume of high-end Ethernet (≥ 200G) ports in China will exceed 6 million, and will maintain a compound growth rate of 45.6% in the future. By 2029, the shipment volume of high-end Ethernet ports in China will exceed 43 million. IDC predicts that hardware spending on generative AI related networks in China will continue to accelerate, increasing from 6.5 billion yuan in 2023 to 33 billion yuan in 2028, with a compound annual growth rate of 38.5%.