Zhipu is the first to disclose its IPO prospectus—could it be aiming for "the world's first large model stock"?

On December 19, Beijing Zhipu Huazhang Technology Co., Ltd. (hereinafter referred to as "Zhipu") officially disclosed the post-hearing prospectus.

Zhipu released its prospectus ahead of Shanghai Xiyu Technology Co., Ltd. (hereinafter referred to as "MiniMax"), the parent company of MiniMax, and is seen by many in the market as having a better chance of becoming the "world's first large model stock".

It has been reported that MiniMax has passed the HKEX hearing, but has not yet released its prospectus.

Since the title of "first stock" is generally determined by the listing date, if Zhipu and MiniMax were to be listed on the Hong Kong Stock Exchange on the same day, it might be hard to say which is the first.

With the disclosure of this prospectus, the public has gained deeper insights into Zhipu's business operations.

From 2022 to 2024, Zhipu's revenues were 57.4 million, 125 million, and 312 million yuan respectively.

Specifically, Zhipu focuses on MaaS (Model as a Service), with revenues divided into local and cloud deployment.

The main source of income is providing customers with private, exclusive AI models, which generated 264 million yuan in 2024, accounting for more than 80%.

However, local deployment is more like a "one-off project."

According to the revenue recognition method, Zhipu provides localized deployment solutions in package form, with pricing based on model type and scale, amount of computing resources included, and implementation costs. Package prices can be charged as a one-time fee or annually.

Generally, each local deployment project commands a higher price, which can be seen in Zhipu's customer structure: the top five clients contributed 142 million yuan in revenue in 2024, accounting for 45.5%.

By contrast, cloud deployment generates revenue in a more "steady and gradual" manner.

Zhipu's cloud deployment business charges based on token consumption contracts, and pricing also depends on subscription duration, model type and scale, and included computing resources. However, this business segment currently accounts for less than 20%.

To support these two delivery models, Zhipu has built a full-stack product matrix, including small edge models, economic models, and flagship large models with tens of billions of parameters, covering various parameter scales to address specific customer needs.

At the same time, this matrix provides dialogue, general agents, code generation, image understanding, text-to-image/video, voice interaction, and other functions, achieving full coverage of various application scenarios.

Currently, Zhipu remains in a state of significant loss, with losses of 144 million, 788 million, and 2.958 billion yuan from 2022 to 2024, totaling 3.89 billion yuan.

The year-on-year increase in losses is mainly driven by computing power service fees incurred during R&D expenditures.

In 2024, Zhipu's computing power service fees reached 1.552 billion yuan, already about five times its revenue for the period.

Zhipu faces considerable pressure coping with losses given its current cash position; as of the end of June 2025, its cash and cash equivalents stood at 2.552 billion yuan.

Whether Zhipu, the first to disclose its prospectus, can take the title of "world's first large model stock" is attracting attention.

Risk Warning and DisclaimerThe market carries risks, and investments should be made with caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situations, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article suit their particular circumstances. Any investment made based on this is at the user's own risk.