China AI Compute Boom Leads to Overcapacity and Rising Service Prices

China’s AI compute sector is experiencing a paradox of soaring demand and severe underutilisation. In April, cloud provider Hangyun Technology raised its high‑end compute service rentals by 28.79 billion yuan – a 79% increase – as chip shortages and long server delivery times drive “training + inference” workloads upward. Industry insiders say the market for premium AI chips remains tight and prices stay high.

At the same time, government‑backed AI data centres suffer from chronic idle capacity. Multiple sources report average GPU utilisation rates of only 20‑30%, with some sites below 10%. Over‑investment of up to a trillion yuan in “intelligent computing” facilities has created a bubble, prompting concerns about asset devaluation, high‑leverage financing, and project “dead‑ends”. Several corporate projects have already stalled or been abandoned, raising the risk of broader financial strain on local governments and taxpayers.