According to a research report, the liquid cooling industry is transitioning from thematic expectations to a stage of order fulfillment and profit realization. Investors are advised to focus on supply chain companies with production capacity flexibility, customer certifications, and capabilities in core components.

The contract manufacturing chain offers significant short-term profit potential. The mass production of platforms like GB200/GB300 is straining the capacity of leading Taiwanese and international manufacturers. Mainland Chinese companies, leveraging their precision machining and assembly capabilities, are well-positioned to capture the resulting overflow of orders.

Supply chains for independent brands exhibit the strongest long-term growth potential. By capitalizing on their certification assets and customer interfaces, these firms are strategically positioned within the ecosystems of Cloud Service Providers (CSPs), Original Design Manufacturers (ODMs), and domestic GPUs. They are currently in the prototype and small-batch validation phase.

Horizontal expansion into new product categories presents opportunities for differentiation. New products such as power distribution systems and liquid cooling for optical modules, while having lower individual value, offer access to a broad customer base and easier market entry.

The report suggests focusing on three main avenues: profit potential in the contract manufacturing chain, certification breakthroughs for independent brands, and volume expansion for new horizontal product categories. The key points are as follows:

AI Computing Infrastructure Enters the Rack-Level Liquid Cooling Era

AI computing infrastructure is evolving from server-level cooling to the era of rack-level liquid cooling. As platforms like GB200, GB300, and Rubin continue to increase power per rack, chip density, network bandwidth, and memory capacity, the thermal bottleneck is no longer just about controlling the temperature of individual GPUs/CPUs. It now involves managing the entire rack's heat load. Consequently, liquid cooling is advancing from single-point cold plate solutions to a foundational infrastructure capability covering chips, power supplies, networking, optical modules, and rack systems.

The NVIDIA Rubin platform achieves 100% liquid cooling for the entire system, prompting cloud providers and data center operators to accelerate their transition to liquid cooling architectures. Simultaneously, the rising power density of custom ASICs developed by leading CSPs like Google, Meta, and Amazon is also driving the adoption of more efficient cooling solutions in custom AI servers.

Liquid Cooling Demand Expands Beyond Core Chips

The demand for liquid cooling is expanding continuously from core computing chips to multiple heat sources across entire server racks. In 2024, the penetration rate of liquid cooling in AI data centers has reached 14%, marking the accelerated adoption phase of liquid cooling for AI data centers. It is projected that the global market for liquid cooling in AI data center servers could reach $21.8 billion by 2027.

Beyond GPUs and CPUs, areas like busbar power distribution systems and optical modules are becoming new application scenarios for liquid cooling. For power distribution, liquid-cooled busbars can significantly enhance a rack's power capacity. At 48V, liquid-cooled racks can achieve up to 750kW, compared to approximately 140kW for air-cooled solutions. The global market for liquid-cooled DC busbar components in AI data centers is forecast to reach about $230 million by 2026, potentially growing to $640 million by 2032.

For optical modules, the power consumption of 800G OSFP has reached 25W-30W, with 1.6T DR8 DSP solutions peaking at 40W-50W. The penetration rate of liquid cooling is expected to rise from around 5% in the 800G era to about 40% in the 1.6T era. The market size for optical module liquid cooling is predicted to exceed $6.3 billion by 2030.

Core Barriers in the Liquid Cooling Supply Chain

The core barriers in the liquid cooling industry chain stem from precision manufacturing, reliability validation, and large-scale delivery capabilities. Products like cold plates, CDUs, manifolds, UQDs, piping, and liquid-cooled racks involve complex processes such as microchannel machining, vacuum brazing, seamless welding, leak testing, and thermal cycling. These processes impose high requirements on equipment investment, process teams, quality control, and customer certification.

The industry is not simply about assembly and capacity expansion. Consistent yield rates and long-term delivery experience determine whether a company can enter the mainstream supply chain. Taiwanese companies maintain a leading position in the NVIDIA and North American CSP liquid cooling supply chains due to their early certifications, deep customer relationships, and overseas production capacity layout.

Global Supply Chain Dynamics and Mainland China's Role

The global liquid cooling supply chain is currently characterized by the dominance of Taiwanese manufacturers, with Mainland Chinese manufacturing accelerating its role in order fulfillment. North America remains the largest region for liquid cooling demand, accounting for approximately 65% of the market share. Southeast Asia, Mainland China, Europe, Japan, and South Korea follow with shares of about 15%, 10%, 5%, and 5%, respectively.

On the supply side, Taiwanese manufacturers still hold a strong advantage in key components like cold plates and CDUs, with an estimated overall market share of 50%-60%. However, as Taiwanese capacity becomes saturated, the outsourcing proportion for mature liquid cooling components has exceeded 50%. Mainland Chinese enterprises, leveraging manufacturing clusters in East and South China, are capturing this capacity overflow in areas like CNC machining, welding, piping, connectors, assembly, and testing. They are also expected to gradually increase their participation in domestic AI computing centers and the domestic GPU ecosystem.

Three Pathways for Mainland Chinese Enterprises

The liquid cooling industry has entered a phase of order and profit realization. Mainland Chinese enterprises are breaking through the Taiwanese-dominated landscape via three parallel paths: independent brands targeting CSP and domestic GPU ecosystems for long-term growth; the contract manufacturing chain absorbing overflow capacity from Taiwanese firms to release short-term profit potential; and horizontal expansion into differentiated products like optical modules and busbars to broaden customer reach. The overall landscape is taking shape with contract manufacturing realizing profits, independent brands accumulating value, and new product categories achieving differentiated breakthroughs.

Key Risk Factors

Potential risks include AI computing demand falling short of expectations; risks associated with technological path iteration and substitution; intensified industry competition leading to declining profitability; risks related to core customer certification and market entry barriers; and geopolitical and overseas policy risks.