Skip to content
The global refrigeration industryHolodPrice catalogue ↗

Everything shaping
the world of cooling.

EN Select language
РусскийEnglish

Markets & companies

Shift to Rack-Level Liquid Cooling Delivery: What Changes in the Thermal Management Supply Chain

According to IDC, revenue of liquid-cooling server suppliers in China grew 153.1% year-on-year in H1 2026, with a projected five-year CAGR of approximately 49%. The key shift is the move from selling individual components to delivering pre-integrated, rack-level systems, which is restructuring roles across the thermal management supply chain.

Illustration: liquid cooling in server rack equipment
制冷快报

The point of cold application is shifting: from the machine room — into the rack. According to an IDC report, revenue of liquid-cooling server suppliers in the Chinese market grew by 153.1% year-on-year in the first half of 2026. A five-year compound annual growth rate of supply volumes is projected to approach 49%. At the same time, the report identifies a less visible but more consequential shift for the industry: high-density compute projects are now being delivered as complete racks rather than individual servers.

Three factors pushing the market toward liquid cooling

The first is compute load density. The growing power requirements for AI training and inference are pushing per-rack thermal loads well beyond what air cooling can still tolerate. Where a rack once consumed on the order of ten kilowatts and an air system could still manage through organized supply and exhaust airflow, further increases in density make the heat-exchange capacity and noise levels of air cooling unsustainable. Liquid cooling is no longer an option.

The second is tightening energy-efficiency requirements. PUE regulations for new large-scale data centres continue to tighten; in regions with hot, humid climates, the energy shortfall of air-based systems is especially acute. Higher heat-exchange efficiency and lower fan power make liquid solutions one of the paths to regulatory compliance.

The third is a change in delivery format. In high-density projects, servers are no longer procured as individual machines for on-site assembly. The rack becomes the unit of delivery: cold plates, distribution manifolds, quick-disconnect couplings, the CDU, and leak-detection systems are integrated at the factory, leaving only connection work on site.

Restructuring the supply chain

In the air-cooling era, the key thermal-management equipment in a data centre was precision air conditioning; competition centred on sensible-heat share, energy efficiency, and supply/exhaust airflow design. The buyer purchased a unit.

Liquid cooling brings the coolant inside the rack and extends the chain: a cold plate sits against the chip, the CDU distributes and removes heat, manifolds and quick-disconnect couplings feed each level, and an external cold source is still required on the outside. According to the trade publication Refrigeration Express (制冷快报), domestic companies operating in the data-centre thermal-management segment are entering the market from different angles — through cold plates, CDUs, manifolds, or external cold sources; manufacturers of compressors, pumps, and heat exchangers cover the component level. Firms previously specialised in machine-room air conditioning are forced to retool their competencies from "blowing the whole room" to "heat exchange and flow management inside the rack."

Competition criteria are changing too: instead of comparing the COP of a single unit, the focus shifts to system-integration capability, leak-risk control, delivery lead times, and the ability to commission an entire rack at the factory before shipment.

Three barriers to rack-level delivery

The first barrier is pre-integration. Cold plates, manifolds, quick-disconnect couplings, and the CDU must be assembled and pressure-tested at the factory; on site, only the final connection remains. This demands that the supplier possess competencies in designing and testing the rack as a whole, not merely shipping components to the site.

The second is interfaces and standards. Specifications for quick-disconnect couplings, manifold materials, and coolant types remain unstandardised. Interface mismatches hinder interchangeability between racks and between suppliers, and turn spare parts into an operational burden.

The third is liability allocation. A rack is assembled from components by different manufacturers; in the event of a leak or thermal-management failure, the question of who bears responsibility is currently resolved primarily through project-contract terms. This is a new challenge for suppliers transitioning from selling a product to delivering a system.

What this means for the industry

153.1% is realised revenue growth; 49% is a five-year forecast; these figures should be considered separately. Liquid-cooling penetration is still concentrated in the high-compute-density segment; medium- and low-density data centres will remain on air cooling in the foreseeable future.

For companies in the cold industry, however, the shift has already happened: the watershed of competition is no longer whether a company has a liquid-cooling product, but whether it can deliver a reliable, rack-level thermal-management system. Those who master pre-integration and keep leak risk on the factory side will gain access to the next stage of the market.

Source: Refrigeration Express (制冷快报), author Kang Yuting, 12.11.2026.

ORIGINAL SOURCE

制冷快报 ↗

Visit the original source for additional details.

← Back to news