New Stocks Preview | "Decentralized" vs. "Centralized", what will Breakthrough rely on in the future to outwit the market surrounded by large factories?

In the past two years, from OpenAI, Anthropic to DeepSeek, from overseas tech giants to domestic internet companies, the rapid improvement of large model capabilities has driven an exponential growth in AI reasoning demand. In this process, a clearer trend is emerging: the future competition in the AI industry is not only a competition of model capabilities, but also a competition of computing power supply efficiency, resource scheduling capabilities, and infrastructure operation capabilities. In this context, FutureAI has submitted an application for listing to the Hong Kong Stock Exchange. As an independent distributed cloud computing service provider, in a high-cost hardware ecosystem dominated by chip giants such as NVIDIA and in a long-term tug-of-war where enterprises are urgently seeking low-cost distributed computing power, FutureAI has opened up a path that is completely different from traditional centralized cloud giants. If we calculate based on the daily average Token consumption in 2025, FutureAI has ranked first among independent AI cloud computing service providers in China. However, on the flip side, the company's core AI services are still in the stage of losses and "burning money" for expansion. This technology company, touted as the "number one independent AI computing power," is weaving a myth supported by a capital bubble, or has truly grasped the long-term profitability gate of the AI Agent era. Growth "engines": from edge computing to computing infrastructure Different from large cloud service providers like Alibaba Cloud and Tencent Cloud, who rely on self-built data centers for expansion, FutureAI initially entered the edge cloud market. Its core idea is to use its self-developed computing power scheduling platform to integrate a large number of dispersed, heterogeneous, and low-utilization computing resources, and then provide services to customers in a unified manner. By the end of 2025, the company's computing network has covered over 4600 computing nodes globally, distributed in more than 1340 counties and cities. The prospectus cited data from Torch Insight Consulting, showing that in terms of the number of computing nodes, the company operates the largest computing network in China. The advantage of this model is that traditional cloud manufacturers often need to invest huge capital in building servers and data centers, while FutureAI plays more of a role in platform scheduling, creating network effects by connecting third-party computing resources, enabling it to achieve network expansion with relatively lower capital expenditure, while also increasing the utilization of idle resources. It is understood that from 2023 to 2025, the company's total revenue achieved 358 million yuan, 558 million yuan, and 770 million yuan respectively. Among them, edge cloud computing services have always been its core source of revenue, with revenue reaching 358 million yuan, 548 million yuan, and 651 million yuan respectively during the same period, accounting for 99.9%, 98.1%, and 84.5% of total revenue. The continuous growth in demand for low-latency computing and content distribution in scenarios such as short videos, live broadcasts, online entertainment, and real-time interactions has provided a stable growth foundation for the company's edge cloud business. In addition, in the AI cloud business, in 2023, the company's AI cloud revenue was only 265 thousand yuan, which was almost negligible. By 2024, this business revenue had grown to 10.387 million yuan; and in 2025, it further soared to 119 million yuan, with a year-on-year growth of over ten times, accounting for 15.5% of the company's total revenue, gradually becoming another major source of revenue for the company. As large models enter the application landing stage, the industry's demand for GPU resources has gradually shifted from the training side to the reasoning side. Compared to the training market which is primarily dominated by large tech companies, the participants in the inference market are more diverse, with many small and medium-sized developers, AI-native companies, and overseas applications requiring flexible and low-cost inference computing power services. FutureAI's business model revolves around this demand. The company provides GPU cloud services, model API services, and model hosting services, helping customers obtain AI inference capabilities without having to purchase GPU devices themselves. At the same time, the company also improves model running efficiency through inference optimization technology. Of particular note is that the company is further extending towards the so-called "Agentic Infrastructure." The management believes that in the future, the main users of cloud computing resources may no longer be human users, but AI Agents with autonomous execution capabilities. In line with this trend, the company has laid out Agent hosting, memory systems, sandbox environments, and tool calling systems, aiming to build a new type of infrastructure platform for the era of intelligent agents. Currently, whether it is OpenAI, Anthropic, or leading domestic model manufacturers, are actively promoting the evolution of Agent capabilities. If AI Agents become widely popular in the future, the importance of underlying computing power scheduling, memory storage, and task execution environments is expected to further increase. However, it should be pointed out that currently the infrastructure for intelligent agents is still in a relatively early stage, with the business model and market size yet to be fully validated, so this business more reflects future growth expectations rather than current performance contributions. Pressure on gross profit under high growth, testing the ability of "self-generation" From the perspective of revenue growth, FutureAI is undoubtedly a company with strong growth potential. From 2023 to 2025, the company's revenue reached 358 million yuan, 558 million yuan, and 770 million yuan respectively, with a two-year compound growth rate of 46.6%. However, when it comes to profitability, the company still faces certain pressures. It is understood that the company's overall gross profit margin decreased from 17.7% in 2023 to 12.3% in 2024, and further to 9.4% in 2025. Among them, the gross profit margin of the edge cloud business decreased from 17.8% to 13%; and the AI cloud business is still at a negative gross profit, with a gross profit margin of -10.7% in 2025. The main reason for this phenomenon is the rapid increase in computing costs. Both edge cloud and AI cloud businesses are essentially resource-intensive industries. As the business scales up, the company needs to procure more computing resources, cloud services, and GPU capabilities, leading to cost growth faster than revenue growth. From 2023 to 2025, the company's net losses reached 189 million yuan, 294 million yuan, and 223 million yuan respectively. Although some of the losses are due to non-operating factors such as fair value changes in convertible preferred shares, even after excluding these impacts, the company's adjusted net losses increased from 37.08 million yuan to 105 million yuan, and the adjusted loss rate increased from 10.3% to 13.6%, indicating a lack of clear profit turning point at the core operational level. As a result, in 2025, the company's net cash outflow from operating activities reached 136 million yuan, higher than the levels of the previous two years. The continued losses combined with cash consumption also make the company highly dependent on external financing. By the end of 2025, the company was still in a net debt position, with a net debt size close to 890 million yuan. From the perspective of industry competition, the domestic AI cloud market is currently heating up rapidly. In addition to major comprehensive cloud vendors like Alibaba Cloud, Tencent Cloud, Huawei Cloud continuing to increase investments in AI infrastructure, platforms like Volcano Engine, Baidu Intelligent Cloud, and many emerging AI cloud platforms are also actively competing for market share. Compared to large cloud manufacturers, FutureAI's advantages lie in its distributed architecture, resource scheduling capabilities, and independent third-party positioning, but there is still a certain gap in brand influence, customer resources, and financial strength. In the future, as the cost of inference continues to decrease and model capabilities tend to become more homogenized, the industry's competitive focus is likely to shift further towards price and scale effects. At that time, whether the company can maintain its technological edge and establish sufficient competitive barriers remains to be seen. For the market, the biggest point of interest in FutureAI is not its current profit levels, but rather whether it can establish a truly network-effect-based infrastructure platform in the future AI reasoning and intelligent agent era. If the company can further transform its current developer ecosystem, computing network scale, and technological capabilities into commercial barriers, it is expected to share in the long-term growth dividends of AI infrastructure; however, if intensified industry competition continues to pressure profitability, its high growth story will also face a reality test.