Zhipu AI Posts 954 Million Yuan in H1 Revenue, Up Nearly 400% YoY
nashnova research
Zhipu posted RMB 954 million in first-half 2026 revenue, up nearly 400% year-on-year, with API services accounting for over 86% — a sign that enterprise adoption of China's homegrown large models is shifting from trial to scaled procurement.
Nearly 400% growth — what does that actually mean?
Zhipu's H1 total revenue reached roughly RMB 954 million (about $142 million), up nearly 400% year-on-year.
This means → H1 2025 revenue was around RMB 200 million, so the top line roughly quintupled in twelve months.
In plain terms = a year ago Zhipu was an AI company with strong tech but modest revenue; now it is generating real commercial income.
Where did the money come from? One business line drove 86%
The open-platform and API business brought in roughly RMB 825 million in H1, up 2,735.7% year-on-year — over 86% of total revenue.
This means → Zhipu's income is heavily concentrated; nearly all growth came from enterprises calling its large-model APIs.
In plain terms = companies are no longer just testing Zhipu's models — they are plugging them into products and workflows, paying by usage on an ongoing basis.
What does the API explosion really tell us?
A 2,735.7% year-on-year jump implies H1 2025 API revenue was under RMB 30 million — an extremely low base.
This reflects two things: first, the API business barely existed a year ago; second, while this year's growth is striking, absolute scale is still early-stage.
In plain terms = the percentage looks dramatic, but going from "almost nothing" to "RMB 825 million" means the market just opened up — it does not mean Zhipu has locked it in.
What to watch in the second half?
First checkpoint: whether API revenue growth can hold up — the base is now much higher, so the rate will inevitably slow; the question is how much.
Second checkpoint: whether non-API businesses (currently only about 14% of revenue) can open new income streams and reduce single-line dependence.
This means → Zhipu's "monetisation story" is only in its opening chapter; H2 numbers will determine whether that story holds.
市场有风险，内容仅供研究参考，不构成投资建议。