Cooling down the chips, this liquid cooling company has raised nearly 100 million yuan in financing.

All electronic devices generate heat. How to efficiently remove this heat in a timely manner directly determines whether the device can operate stably in the long term. In the past few decades, air - cooling, which relies on air flow for heat dissipation, has always been the mainstream. Liquid - cooling is more commonly used in scenarios where extremely high - intensity computing tasks are processed, mainly concentrated in high - end fields such as supercomputing.

With the explosive growth of AI computing power demand, the power density of data centers has been rapidly increased. The physical limit of traditional air - cooling is being approached, and liquid - cooling is starting to transition from a "high - end special configuration" to a "large - scale infrastructure". The change in the computing power architecture is opening up an unprecedented industrial space for liquid - cooling technology. In the industrial sector, 2026 is generally regarded as the key turning point for liquid - cooling heat dissipation to move from the "market verification period" to the "rapid growth period".

Recently, Shenzhen Green Cloud Atlas Technology Co., Ltd. completed a 100 - million - yuan Series A financing round, led by Yunze Capital and jointly funded by two other listed companies. The funds from this round will mainly be used to supplement working capital, expand the sales and technical teams, and promote the construction and expansion of production lines in places such as Suqian.

As early as August 2025, Yunze Capital led a 100 - million - yuan Pre - Series A financing for Green Cloud Atlas. They believe that 2026 will be the real "explosion year" for liquid - cooling in China.

From a policy perspective, with the accelerated construction of AI intelligent computing centers, the single - card power consumption of high - power chips (such as H100, B200, etc.) has increased to the range of 700W - 1000W. Traditional air - cooling methods can no longer ensure the stable operation of chips under full load, and liquid - cooling is gradually becoming a necessary condition for performance release. At the same time, first - tier cities are increasingly strict in regulating the PUE indicators of data centers, generally requiring them to approach below 1.2. The liquid - cooling solution can significantly reduce the PUE from over 1.5, and achieve a structural optimization of operating costs by saving electricity costs in the long term.

From a technical path perspective, currently, cold - plate liquid - cooling is still the mainstream in the market. However, as the chip power continues to rise and the power density of single cabinets continuously breaks through the upper limit, from 2028 to 2030, the industry will enter a window period for large - scale switching from cold - plate to immersion liquid - cooling. Currently, a large number of mature manufacturers have gathered in the cold - plate technology field. In contrast, immersion liquid - cooling is still in the early stage of engineering and large - scale implementation.

Of course, technological breakthroughs alone are not enough; the economics also need to make sense. Yunze Capital said that in 2024, there was a key change in the cost of cooling liquid. It gradually shifted from the fluorinated liquid system, which used to cost hundreds of thousands of yuan per ton, to the synthetic oil system in the range of tens of thousands of yuan. This means that for the first time, liquid - cooling is truly "economically viable" for large - scale commercialization.

The downward shift of the cost curve has made liquid - cooling move from being technically feasible to economically feasible.

Nearly 200 million yuan in orders in a year

Like most liquid - cooling companies, Green Cloud Atlas was founded by industry veterans.

Hu Shixuan, the general manager of Green Cloud Atlas, once served as the vice - president of listed company Wangsu Science & Technology. Wangsu Science & Technology was established in January 2000 and is a computing service provider mainly offering services such as CDN, cloud computing, cloud security, and global distributed data centers. Green Cloud Atlas was incubated within Wangsu Science & Technology in 2011 and became independent in 2015. It focuses on providing overall solutions for green computing power centers based on immersion and cold - plate liquid - cooling technologies.

In the niche liquid - cooling segment, Green Cloud Atlas has accumulated more than 10 years of R & D experience and real - world scenario data. Its products and systems can cover various application forms such as data centers, edge computing, and container liquid - cooling, and a number of large - scale implementation cases have been formed. The target customer group of Green Cloud Atlas highly overlaps with the core large customers of Wangsu Science & Technology's original CDN computer rooms, and it naturally inherits a mature and stable downstream channel resource system.

The economic advantages of liquid - cooling technology in overseas markets have gradually emerged, especially in regions with high temperatures and high electricity prices. Hu Shixuan took the computing power center of a well - ranked global university in Hong Kong as an example. The center's annual electricity bill is about 200 - 300 million Hong Kong dollars, and it currently mainly relies on an air - cooling system. "If the entire system is switched to a liquid - cooling solution, about 20% of the power cost can be saved, which means a net savings of 50 - 60 million Hong Kong dollars per year."

From the perspective of industrial cooperation, Hu Shixuan revealed that the company has established a technology mutual - recognition mechanism with first - tier manufacturers such as Intel and Dell and is gradually integrating into the international computing power infrastructure ecosystem. He expects that in the next 2 - 4 years, the company is expected to further break through the cooperation threshold with core manufacturers such as NVIDIA and AMD. Once the system - level connection is completed, the business growth curve will show an obvious geometric amplification effect.

In 2026, the scale of Green Cloud Atlas' on - hand orders was close to 200 million yuan. Hu Shixuan believes that with the concentrated implementation of several large - scale computing power node projects in China, the company's revenue in 2026 is expected to reach the range of 300 - 400 million yuan, officially entering a new stage of large - scale development.

"In the past, we had to repeatedly explain to customers whether it was necessary to introduce liquid - cooling; now, more customers come to us actively to discuss solutions." Hu Shixuan said that the focus of customers has shifted from "whether to use liquid - cooling" to "how to allocate the proportion of liquid - cooling", whether it is 70% or 30%. This means that liquid - cooling is no longer a technical option in the concept verification stage but has entered the practical stage of engineering configuration, with clear demand boundaries and direct and explicit order intentions.

The explosion of the 100 - billion - yuan track

Driven by the wave of AI computing power, liquid - cooling is rapidly transitioning from a relatively niche infrastructure technology to a core track with a trillion - yuan imagination space.

A research report released by Goldman Sachs pointed out that the global server cooling market is facing a structural growth opportunity. As the expected shipments of AI servers continue to be revised upwards and the penetration rate of liquid - cooling technology accelerates, it is expected that from 2025 to 2027, the global server cooling market will achieve year - on - year growth of 111%, 77%, and 26% respectively, and the overall market size will reach 17.6 billion US dollars by 2027. This means that liquid - cooling is no longer just "computing power support" but is becoming a key infrastructure for supporting the operation of the new - generation computing power system.

In this context, the competition logic of the liquid - cooling industry has also changed. Platform providers are increasingly dominant in terms of technical specifications and procurement systems. The development rhythm is accelerating, and the upfront R & D and delivery investment has increased significantly, causing the overall industry entry threshold to continue to rise. Whether a company can participate in platform - level design, undertake the rapidly increasing demand, and maintain a reasonable gross profit level in the fierce price competition is becoming the core indicator for measuring the comprehensive strength of a liquid - cooling enterprise.

2026 is regarded as both the "first year" of the rapid growth of liquid - cooling and the watershed for the reshuffle of the industry competition pattern. Some enterprises will complete the transition from engineering suppliers to platform - type capability providers, while others may be confined to the low - gross - profit range of standardized manufacturing.

The industry generally believes that the value proportion of liquid - cooling in the overall BOM (Bill of Materials) of servers will continue to rise in the next few years. As the computing power density continues to increase, the heat dissipation system will gradually evolve from a "cost item" to a "key fulcrum for performance release and stable operation", and its industrial status will be significantly improved.

Under this trend, the entire industrial chain structure will inevitably change. The binding degree between liquid - cooling solutions and computing power platform specifications is constantly deepening, and chip manufacturers and cloud service platforms are more deeply involved in the process of defining the heat dissipation architecture. The competition focus among suppliers will shift from simple "manufacturing capabilities" to "design participation". Those who can participate in the joint design at the early stage of the platform architecture and take the lead in mastering the first - hand technical specifications and system standards are more likely to gain a long - term advantage in future large - scale procurement.

This means that the liquid - cooling industry is entering a new cycle of "competing for platform discourse power" from the stage of "competing for delivery capabilities". Green Cloud Atlas chooses to lay out in "next - generation technology" in advance. Its core goal is not to seize the market in the short term but to build technical barriers and comprehensive strength under future scale effects by accumulating operating data, engineering experience, and delivery capabilities in advance.

This article is from the WeChat official account "Dongshisi Tiao Capital" (ID: DsstCapital), author: Wei Xianghui, editor: Liu Yanqiu, published by 36Kr with authorization.