Second time submitting to the Hong Kong Stock Exchange, PaoCloud aims to leverage the AI inference boom to retell its IPO story.
```
Author | Huang Yu
Amid the AI infrastructure boom, Paiou Cloud is unwilling to easily give up the opportunity to make a splash in the capital market.
On June 10, Pai Xiang Future Group (hereinafter referred to as "Paiou Cloud") once again submitted a listing application to the Hong Kong Stock Exchange.
This is not Paiou Cloud's first attempt at the capital market. As early as June 2025, Paiou Cloud had submitted a prospectus to the Hong Kong Stock Exchange, but it expired due to failing the hearing. A year later, standing again before the HKEX, Paiou Cloud’s joint sponsors remain ICBC International and Shenwan Hongyuan Hong Kong.
If last year when submitting the form, Paiou Cloud was still telling the story of being "China’s largest independent edge cloud service provider," now, Paiou Cloud further strengthens its AI narrative in the prospectus, and its AI cloud computing services are beginning to contribute more substantial revenue.
As a platform enterprise connecting GPU resources and AI developers, Paiou Cloud obviously hopes to leverage this round of industry wave to switch its own valuation logic.
But for investors, the more noteworthy question is not whether the AI market is large enough, but whether Paiou Cloud is truly an AI cloud platform with technical barriers, or a “computing intermediary” relying on integrating compute resources for growth. This issue may, more than the IPO itself, determine its future performance in the capital market.
How does Paiou Cloud make money?
Paiou Cloud’s current position in the AI industry chain is not easy to understand.
It’s neither a model company like OpenAI or Anthropic, nor a cloud provider with massive data centers like Alibaba Cloud or Tencent Cloud, but more like an intermediary platform connecting compute supply and demand.
Simply put, Paiou Cloud’s core business model is to integrate globally dispersed idle computing resources, and then provide computing services to customers via a unified scheduling platform.
Paiou Cloud first ventured into the edge cloud market.
In recent years, as live streaming, short videos, real-time interaction and other services rapidly grew, many applications demanded higher network latency and computational efficiency. Paiou Cloud connected telecom operator data centers, regional data centers and various edge nodes to provide edge computing and content distribution services for clients, thereby establishing a distributed computing network covering the country.
By the end of 2025, Paiou Cloud had operated over 4,600 computing nodes, covering more than 1,340 cities and counties worldwide.
What truly changed the company’s direction, however, was the generative AI wave.
As large model applications became commercially viable, demand for GPU computing and inference services surged. Paiou Cloud began shifting its business focus to AI cloud computing.
Currently, Paiou Cloud has formed a three-layer business system.
The lowest layer is the IaaS infrastructure layer, responsible for integrating GPU and edge computing resources; the middle layer is the MaaS (Model as a Service) platform, providing model API calls and inference services for clients; the upper layer is Paiou Cloud’s key focus, Agent Infrastructure, offering agent hosting, memory systems, tool invocation, and sandbox environment capabilities.
In other words, Paiou Cloud does not engage in basic large model R&D, but hopes to become the foundational infrastructure platform connecting compute, models and developers in the AI era.
Over the past year, the large model industry underwent a major shift—the industry focus moved from training to inference.
As more enterprises embed large models in office, search, customer service, marketing, programming, and other scenarios, inference computing is increasingly becoming a new infrastructure requirement.
Paiou Cloud has clearly become a beneficiary of this trend.
The prospectus shows that in 2024, Paiou Cloud's AI cloud computing revenue was only 10.387 million RMB, accounting for 1.9% of total revenue, with a negative gross margin (gross loss 95.1%); edge cloud revenue accounted for 98.1%.
By 2025, Paiou Cloud’s AI cloud computing service revenue reached 119 million RMB, surging more than tenfold year-on-year, directly raising the proportion of total revenue to 15.5%;
Meanwhile, the number of registered developers on the platform continuously increased, and the average daily token consumption also grew multiple times.
According to analysis from China Insights Consultancy, by April 2026, Paiou Cloud's daily token consumption reached 10,280 billion times, becoming China’s largest independent AI cloud computing service provider.
For the capital market, this means the company has found a second growth curve beyond traditional edge cloud business.
But on the other hand, it indicates investors need to reevaluate the company's valuation logic.
Last year the market saw an edge cloud company; this year, Paiou Cloud hopes to convince the market it is an AI infrastructure enterprise. The valuation systems for these two identities are completely different.
Technology Platform or “Compute Porter”?
However, rapid growth does not mean the business model has already been validated.
From financial data, Paiou Cloud is still in the stage of "revenue increases without profit growth."
The prospectus shows that from 2023 to 2025, Paiou Cloud's revenue was 358 million, 558 million and 770 million RMB respectively, maintaining rapid growth.
Meanwhile, Paiou Cloud has not yet achieved profitability. In 2023, 2024, and 2025, its net losses were 189 million, 294 million and 223 million RMB respectively.
Even more noteworthy is the change in gross margins. From 2023 to 2025, Paiou Cloud's overall gross margin declined not only with the fastest-growing AI cloud business still losing money, but also with edge cloud computing margins decreasing year by year.
This reflects the fierce competitive environment of the cloud computing services market.
Over the past year, as open-source models like DeepSeek rapidly spread, large model call costs continued to fall. At the same time, Alibaba Cloud, Tencent Cloud, Volcano Engine, Baidu AI Cloud and other large cloud providers kept launching price wars, while new AI cloud companies flooded into the market.
Against this backdrop, the market is no longer concerned only with revenue growth speed, but with when Paiou Cloud can achieve profitability.
For the capital market, revenue growth can support short-term valuation, but long-term value still depends on profitability.
If AI cloud business cannot generate positive profit, then no matter how fast the growth, it will hardly gain long-term market recognition.
In its latest prospectus, Paiou Cloud regards technical capability as its core competitiveness.
Disclosed in the prospectus, in 2025, Paiou Cloud's average GPU utilization exceeded 75%, significantly higher than the industry average of 40%-50%.
According to Paiou Cloud, its main advantage is global distributed scheduling capability.
Because different regions have time zone and demand differences, Paiou Cloud can dynamically deploy idle GPU resources to markets with higher demand, thereby improving overall utilization efficiency.
In addition, Paiou Cloud also reduces customer use costs through inference optimization, model optimization and operator optimization.
In the current context of tight compute resources and high compute costs, this is indeed an attractive business model.
Nevertheless, for Paiou Cloud, whether it’s acquiring more GPU resources, expanding overseas node networks, building agent infrastructure or investing in R&D, all require continuous capital input.
Now is a critical period for Paiou Cloud’s transition from edge cloud to AI cloud.
If it can go public during the high-growth phase in AI cloud business, Paiou Cloud will have the opportunity to capture a higher valuation and greater financing capacity, preparing resources for the next stage of competition.
Therefore, this IPO is not just a capital raising exercise, but also an exam for Paiou Cloud's future positioning in the capital market.
After listing, Paiou Cloud will need to further prove that the identity change brought by AI is not just reflected in the prospectus narrative, but is truly visible in revenue structure, profitability, and technical barriers.
Next, the capital market may focus on several key indicators: whether the AI cloud business revenue share can continue to rise, whether GPU utilization can remain high, whether gross margins will improve, and whether agent infrastructure business can form a new growth curve.
Risk Warning and DisclaimerThe market carries risks, investment needs caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views or conclusions in this article fit their own circumstances. Invest at your own risk. ```