You think you're scrolling Douyin (抖音), but Douyin already knows where you'll stop

Token boom, but what does it buy?

It has been reported that China’s National Data Administration disclosed an average daily token call volume exceeding 140 trillion — a headline-grabbing figure that quickly turned into a symbol of China’s AI momentum. Tokens — the smallest processing units for large language models — have become a measurable commodity. Vendors such as OpenAI and Anthropic price access by token use, and Chinese providers have driven prices down so aggressively that capabilities once reserved for large firms are now available to individual developers. Does raw volume mean control? Not necessarily.

Infrastructure, price and concentrated demand

China’s surge is rooted in more than software. The “East Data West Compute” (东数西算) initiative is turning low-cost western power into large-scale data-centre capacity callable by eastern hubs and beyond, and firms like Alibaba Cloud (阿里云), Baidu (百度) and DeepSeek (深度求索) have cut token prices into unprecedented ranges. That engineering+cost advantage has translated into massive local demand: apps such as Douyin (抖音) and platforms run by ByteDance (字节跳动), Meitu (美图) and others generate enormous token traffic. But much of that traffic is naturally concentrated inside China because direct commercial use of OpenAI, Anthropic and Google models is limited on the mainland — a structural effect that amplifies domestic token counts.

Scale without equivalent revenue or rule‑making power

A Feifan Research (非凡产研) analysis reportedly shows Chinese companies account for about 46% of global AI monthly active usage but capture only a sliver of recurring revenue compared with U.S. leaders like OpenAI and Anthropic. The paradox is clear: scale has been gained, but pricing power and ecosystem governance have lagged. Many Chinese providers intentionally mimic OpenAI’s API patterns to ease developer migration — a short-term growth tactic that lowers friction, but one that also cedes long‑term control over standards and developer mindset. Meanwhile, chips and high-end compute remain concentrated in suppliers such as NVIDIA, and export controls and geopolitics continue to shape who can scale top‑tier models.

What comes next?

If tokens answer the “how many” question, the strategic battle is about “which” tokens matter. China’s path forward requires three shifts: convert cost and scale into truly irreplaceable abilities in high‑value domains (finance, healthcare, industrial), move from API compatibility to standard‑setting, and sustain bottom‑level innovation in chips and core algorithms despite geopolitical pressure. The 140 trillion figure is real and impressive. But without a deeper structural leap, it risks remaining an index of activity rather than a proxy for sustained industrial leadership.