Author｜Yang Licheng

Editor｜Chen Xiaoran

On June 25, Midea Group (000333.SZ) announced that its Shunde liquid cooling smart manufacturing base is expected to commence operations in August 2027, with a total investment exceeding 1 billion yuan, accelerating its transition from a home appliance giant to a key supplier of AI computing power infrastructure.

Midea Group disclosed a critical timeline for its liquid cooling strategy during a roadshow event. The Shunde liquid cooling smart manufacturing base, with a total investment exceeding 1 billion yuan, initiated construction in March 2026. Once operational, it will achieve independent large-scale production of core liquid cooling components such as natural cooling magnetic levitation chillers and CDUs (Cooling Distribution Units).

Half a month earlier, Midea signed a strategic cooperation agreement with Alibaba to upgrade their deep collaboration in the AIDC (AI Data Center) sector, jointly developing liquid cooling temperature control solutions for high-load AI clusters.

Fang Hongbo, Chairman of Midea Group, and Wu Yongming, CEO of Alibaba, jointly attended the signing ceremony, underscoring the strategic importance both parties place on this collaboration.

As of the market close on June 25, Midea Group's A-shares traded at 76.50 yuan, down slightly by 0.17%, with a total market capitalization of approximately 582.4 billion yuan and a dynamic price-to-earnings ratio of around 13.18 times.

Path Selection

In the data center liquid cooling sector, Midea and Gree have taken markedly different approaches.

Gree has opted for a "scenario-first" strategy, partnering with IDC operators like SinoNet to refine equipment and engineering capabilities in actual data center environments.

Midea Group, in contrast, focuses on a "product-driven" approach, prioritizing the development of the most technically challenging end-stage core components—CDUs—before expanding upstream to cold sources and auxiliary air cooling to create a standardized product portfolio.

In 2025, Midea launched its liquid cooling core component, the CDU. By 2026, it further introduced a data center thermal management solution integrating chillers, CDUs, air walls, and intelligent control systems. Its magnetic levitation active CDUs combine cold sources and distribution units into one, reportedly saving over 50% of floor space.

This path has been validated in Midea's proprietary projects.

The completed Phase A of the Gui'an Midea Cloud Data Center, housing core computing deployments such as the "Home Brain" and "Factory Brain," utilizes Midea's self-developed AHU direct ventilation system and high-efficiency chillers, achieving a measured PUE below 1.17 and demonstrating a mature core equipment adaptation system.

The under-construction Phase B will deploy liquid-cooled cabinets specifically designed for AI computing power, adopting a hybrid "dynamic dual cold source + liquid cooling" air-liquid cooling solution.

The collaboration with Alibaba provides Midea with a critical springboard to transition from "products" to "systems." Alibaba Cloud, with China's largest public cloud infrastructure and extensive liquid cooling operational experience, can offer Midea real-world scenario validation and demand-driven insights.

Cross-Industry Rationale

The near-simultaneous entry of Midea and Gree, two home appliance giants with market capitalizations exceeding 100 billion yuan, into the liquid cooling sector is no coincidence.

According to CCID Consulting, China's liquid-cooled data center market reached 15.98 billion yuan in 2025, up 45.2% year-on-year, and is projected to surge to 47.04 billion yuan by 2028, more than doubling in size.

More critically, as chip power consumption continues to rise, liquid cooling is transitioning from an "optional" to a "mandatory" solution.

China Business Industry Research Institute forecasts that liquid-cooled server penetration in China will jump to 37% in 2026 and exceed 50% in 2027.

The rationale for home appliance giants entering this space is clear: cold source technologies, manufacturing supply chains, and channel capabilities from the central air conditioning sector can partially transfer to the data center temperature control market.

However, the core barriers to liquid cooling extend far beyond manufacturing, with four critical thresholds:

Design authority through early-stage intervention (entering joint design systems with server manufacturers and top-tier clients), secondary-side control capabilities (pump valves and supply temperatures must deeply integrate with IT loads), full lifecycle data barriers (issues like corrosion and seal failures require years of operational data to surface), and clear delivery responsibility boundaries (daring to guarantee system performance).

From the current landscape, Midea is well-positioned with clear product boundaries, enabling standardized sales of core components like CDUs. Its strategic collaboration with Alibaba provides a closed-loop for computing scenarios and data feedback.

Nevertheless, challenges persist. While CDUs can be standardized, cold plates, quick connectors, and entire cabinet interfaces remain highly dependent on server platform definitions. Product launches do not equate to mastering server-side definition rights.

Structural Challenges

Beneath the high-growth narrative of the liquid cooling industry lies a structural challenge that is unlikely to resolve shortly: a significant disconnect between the industry's overall development pace and capital markets' optimistic projections.

CCID Consulting reports that in 2025, most domestic enterprises engaged in liquid cooling businesses failed to meet the growth expectations set by earlier market forecasts, plunging the entire sector into a classic development mismatch dilemma.

Specifically, downstream cloud providers and smart computing parks released substantial technology procurement intentions and project bidding plans based on AI computing capacity expansion plans, rapidly heating up industry expectations. However, multiple objective factors—including data center construction approval processes, overall computer room (data center) renovation timelines, and upstream-downstream hardware-software compatibility testing—delayed substantive project launches, equipment installations, and revenue recognition, creating a cyclical mismatch where "demand expectations lead, while business operations and revenue recognition lag."

Currently, most industry players can only secure small-scale pilot orders and demonstrate projects, with scaling (large-scale) revenue realization pending downstream infrastructure completion and acceptance.

Based on industry project timelines, the critical window for concentrated revenue release across the sector will likely fall between 2027 and 2028. Midea's Shunde liquid cooling smart manufacturing base, scheduled to commence operations in August 2027, aligns precisely with the industry's revenue realization cycle, potentially enabling it to fill supply gaps and capture market share during the industry's growth surge.

Beyond cyclical timing discrepancies, the industry faces significant uncertainties in technology selection and commercial implementation.

Technologically, the two mainstream liquid cooling approaches—cold plate and immersion cooling—continue to refine their application scenarios and cost models. Data centers with varying computing densities, geographical environments, and operational conditions face trade-offs between different adaptation solutions, and the industry has yet to establish unified technical standards, exposing enterprises to long-term R&D investment risks.

Commercially, Midea's strategic collaboration with Alibaba currently focuses on joint technology development and scenario testing, without disclosing binding commercial agreements such as long-term exclusive procurement contracts or annual procurement frameworks.

Whether these collaboration outcomes can translate into steady, large-volume product orders—and thus stable revenue for Midea's liquid cooling business—requires long-term validation.

Viewing the industry's full picture, short-term hype cycles and technological route competitions will not alter the long-term trend of computing power cooling upgrades. However, the industry's competitive logic is undergoing a fundamental shift.

The previous model of gaining market share through single-product manufacturing and low-price distribution is gradually losing effectiveness. Enterprises must undergo a profound transformation, evolving from mere hardware suppliers to computing power infrastructure partners capable of providing end-to-end services encompassing design, commissioning, and operational maintenance.

In this prolonged transformation race, only market participants that successfully integrate large-scale manufacturing, scenario adaptation, and full lifecycle operational maintenance capabilities will deeply embed themselves in the AI computing infrastructure ecosystem and seize the core development opportunities presented by industrial upgrading.