Cost analysis and value chain analysis data centre cooling

By Yulin Wang, Senior Technology Analyst at IDTechEx

On the chip and server level, cooling components include cold plate systems (e.g., cold plates, hoses, fluid distribution network), vapor chambers, and fans. On rack and facility level, cooling components include manifolds, coolant distribution units (CDUs), fan doors, rear door heat exchanger (RDHx), computer room air controllers (CRAC), and computer room air handlers (CRAH).

More details on different cooling components are included in IDTechEx’s report, ‘Thermal Management for Data Centers 2025-2035: Technologies, Markets, and Opportunities’.

Thermal costs analysis

Depending on the components, configurations, purchase volume, customer relationships, and materials, the costs of components would vary significantly.

Thermal costs analysis: components on chip and server, as well as rack and facility level. Source: IDTechEx

Copper is commonly used for GPU cooling in data centre servers due to the risk of galvanic corrosion associated with aluminium. The cost of direct-to-chip (D2C) liquid cooling systems for GPU servers, normalised per chip, ranges from $200 to $400. T2 red copper is typically preferred for high-performance GPUs because of its superior thermal conductivity compared to brass. 

Cold plates for CPUs are generally at the lower end of this price range, while those for GPUs are at the higher end. These costs include not only the cold plates but also additional components such as hoses, fittings, fluid distribution manifolds within the server, and quick disconnects (QDs). A key cost driver is the number of QDs, which are particularly expensive, especially at larger diameters. While cold plates themselves are relatively low-cost, the value and expense of the cooling system lie in the complete assembly. This integrated system ensures efficient heat dissipation and operational reliability, making the investment in high-quality components worthwhile.

Data centre cooling value chain

The data centre cooling value chain is long and complicated, requiring collaboration among multiple stakeholders. To simplify this value chain, IDTechEx divides the players in data centres into several key sections. Ultimately, decision-making is based on the collaboration between players in the supply chain.

At the server and chip level, such as cold plate installation, extensive collaboration occurs between chip manufacturers and server manufacturers. Chip manufacturers provide the form factors of their chips, which guide cold plate suppliers in manufacturing cold plates. Typically, chip and server manufacturers work with multiple cold plate vendors. Once the cold plates are installed, they perform internal testing to identify the cold plate that offers optimal thermal performance. During this testing process, IDTechEx believes that if server manufacturers install cold plates upfront, they are more likely to hold the decision-making power, although thermal performance also needs to be evaluated by chip manufacturers. However, in many cases, it also depends on the customers (data centre users). If they prefer to use server boards and install cold plates specified by themselves, the decision-making power ultimately goes to data centre owners.

For cooling solution providers, such as cold plate vendors and immersion cooling solution vendors, the primary customers are server OEMs, chip makers, and system integrators rather than data centre end-users. These providers often face limited negotiation power compared to larger entities in the supply chain, such as server OEMs, chip makers, system integrators, and data centre operators, due to their relatively smaller market influence and the critical role of their customers in dictating specifications and pricing.

Value chain of data centres. Source: IDTechEx

Despite this imbalance, a notable trend identified by IDTechEx is the increasing collaboration across different parts of the supply chain. This shift is driven by the growing complexity of cooling requirements for advanced hardware, such as high-performance GPUs and AI servers. Collaborative efforts aim to align the design, manufacturing, and integration processes to enhance efficiency, reduce costs, and improve the performance of cooling solutions. For example, closer partnerships between cooling solution providers and chip makers can result in bespoke cooling designs optimised for specific processors. Similarly, working with system integrators and server OEMs helps streamline the integration of these solutions into larger systems, fostering innovation and ensuring reliability in high-demand applications like AI and cloud computing. The ‘Thermal Management for Data Centers 2025-2035: Technologies, Markets, and Opportunities’ report from IDTechEx includes more analysis on how the strengthening collaborations would bring opportunities.