In March 2026, Midea launched the construction of its Shunde Liquid Cooling Intelligent Manufacturing Base. According to the company's disclosure, the total investment in this base exceeds CNY 1 billion. It is planned to produce natural cooling magnetic levitation chillers, CDUs, and other products, with production expected to commence in August 2027.
Upon seeing this news, many people's first reaction might be: here comes another home appliance giant chasing the AI boom.
However, after reviewing the annual reports, the story is not that simple.
This is not Midea's first foray into data centers. The company's 2026 semi-annual report shows that Midea Building Technology has long been providing temperature control products such as chiller units and computer room air conditioners to China Telecom, China Mobile, and China Unicom.
Thus, a question more worth asking than "home appliance giant crossing over" emerges: Since Midea has been cooling data centers for years, why continue to move deeper and manufacture a CDU that is closer to the servers?
Let's start with the conclusion: This is not a crossover starting from scratch. Instead, customers' cooling methods have changed, and Midea is extending its existing cold source business towards the server side. What it aims to secure is not just the sale of a single CDU, but a more complete data center temperature control delivery. Public information currently only proves that the products have been implemented and capacity is under construction; it cannot yet prove that this business has been realized on a large scale.
This Is Not a Leap from the Living Room to the Server Room, but a Step Deeper Along the Water Circuit
Let's first clarify a concept that is easily confused.
Chiller units, computer room air conditioners, and CDUs are all related to "cooling," but they occupy different positions. A chiller unit is more like a refrigeration station responsible for producing cooling capacity; a computer room air conditioner mainly handles the server room environment; a CDU, on the other hand, is located between the server room cooling system and the server liquid cooling loop, closer to the servers.
Therefore, Midea's manufacturing of CDUs is not a zero-to-one intrusion into a completely unfamiliar industry. More accurately, it is taking another step towards the server side along the heat dissipation chain, building on its existing cold source, air conditioning, and data center customers.
This step was already written into the annual report in 2025. Midea disclosed that its Building Technology launched the liquid cooling key component CDU that year and jointly developed a new generation of products with leading industry customers.
Here is an important distinction: In the past, Midea mostly solved "how the server room obtains cooling capacity"; with the CDU, it has begun to participate in "how cooling capacity is safely and stably delivered near the servers."
What Exactly Does a CDU Do? It Can Be Seen as a "Transfer Station" Between Two Water Circuits
The full name of CDU is Coolant Distribution Unit.
Why do data centers need it? The technical documents of the Open Compute Project (OCP) provide a direct background: as processor power and power density increase, traditional air cooling cannot always solve the heat dissipation problems of high-performance computing equipment, leading to more widespread application of liquid cooling.
After adopting liquid cooling, the server room can be roughly understood to have two connected but non-mixing water circuits:
One belongs to the server room side, delivering the cooling capacity;
The other belongs to the server side, allowing the coolant to circulate inside the racks and servers, taking away the heat generated by the chips.
The CDU stands between the two. The OCP defines it as a heat exchange device: on the one hand, it isolates the circuits on the server room side and the IT equipment side; on the other hand, it controls the flow, pressure, and temperature of the coolant on the server side; the equipment also monitors anomalies and issues alarms through sensors and controllers.
In less technical terms: The chiller answers "where the cold comes from," while the CDU answers "how the cold is steadily delivered."
It looks like a cabinet equipped with heat exchangers, water pumps, valves, and control systems, but its tasks are not light. Servers need to run continuously for long periods, and liquid leaks, pump failures, temperature fluctuations, or inconvenient maintenance can all bring risks. Therefore, OCP documents list pump redundancy, material and coolant compatibility, sensing and alarming, and maintainability as important design issues.
This also explains why HVAC enterprises are eyeing CDUs: heat exchange, pumps, control, manufacturing, and engineering delivery are all capabilities they are familiar with.
However, "similar capabilities" does not mean "ready to use out of the box." The server side has more stringent requirements for reliability, materials, control, and non-stop maintenance. Midea's choice to jointly develop with leading customers also indicates that the products still need to undergo integration and validation in customer scenarios.
Already Having Chillers, Why Still Insist on Making CDUs?
The answer is not just "selling one more piece of equipment."
First, customers' cooling methods are changing.
As high-power servers become more prevalent, heat dissipation in some server rooms is shifting from pure air cooling to a combination of air and liquid cooling. Suppliers originally only responsible for the server room's cold source, if they stop at the chiller end, can only participate up to the server room side; after customers migrate to liquid cooling, the CDU becomes the key node connecting the original cooling system and the server liquid cooling loop.
Second, what Midea wants to deliver is not just a single piece of equipment.
Investor relations activity records from June 2026 show that Midea describes its data center temperature control business as a service covering consulting, design, equipment selection, product delivery, and post-operation maintenance, extending its product range to cold sources, CDUs, liquid cooling terminals, and digital operation and maintenance.
Based on these public actions, a reasonable judgment can be made: Manufacturing CDUs has both a "defensive" side and an "offensive" side.
"Defensive" means keeping up with the technological changes of existing customers. As customers move from air cooling to liquid cooling, Midea does not want its capabilities to remain solely at chillers and computer room air conditioners.
"Offensive" means expanding its scope of participation in a project. If cold sources, CDUs, terminals, and control systems can be designed and delivered together, the supplier has the opportunity to take on more system coordination work and get closer to the core operational needs of customers.
However, it must be emphasized that this paragraph is an analysis based on product and business layout, not a revenue conclusion already disclosed by the company. Whether participating in more links can truly bring more orders and higher profits still depends on contracts and financial data.
Looking at it from a broader perspective, the CDU also aligns with Midea's enterprise business direction in recent years. In September 2026, Midea stated at its earnings briefing that consumer-facing business is the "moat" for current development, while enterprise-facing business is the "future guarantee to cross cycles," and listed data center liquid cooling as an enterprise business area for firm investment.
Therefore, the CDU is not an isolated new product. It is more like a puzzle piece for Midea to continue pushing its HVAC, manufacturing, and engineering capabilities to enterprise customers.
Is Midea Seriously Making It, or Just Taking a Spot First?
To judge whether a company has truly entered a new business, one cannot just look at press conferences; at least, one must see if products, testing, projects, and capacity are connected.
From public information, Midea has already taken the following four sets of actions:
Product: Launched the CDU in 2025 and jointly developed a new generation of products with leading customers. The company claims that the cooling density of the new generation product has increased by nearly 50% compared to the first generation, while also improving reliability, online maintenance, and heat exchange capacity.
Testing: The 2026 semi-annual report disclosed that Midea has built a comprehensive CDU testing platform equipped with a 1500kW electric heating simulated load, capable of performing single-machine performance and secondary-side loop network group control verification.
Projects: The company disclosed that it has provided CDU products for a T3-level data center in Wuhu; magnetic levitation supplementary cooling CDUs have also achieved batch application in a hyperscale data center project in Malaysia.
Capacity: The Shunde Liquid Cooling Intelligent Manufacturing Base has started construction, planning to form the R&D, manufacturing, and delivery capabilities for liquid cooling products.
Connecting these four things, it can at least be shown that: Midea's CDU has not just stayed at the concept or prototype stage; public information already contains expressions of actual projects and batch applications.
But the boundaries must also be made clear.
First, "total investment exceeding CNY 1 billion" refers to the entire liquid cooling intelligent manufacturing base, not just the CDU. The base also includes products such as magnetic levitation chillers and energy storage temperature control; this money cannot all be counted as investment in the CDU.
Second, entering a supplier whitelist does not equal securing orders. The semi-annual report mentioned that Midea has entered the liquid cooling supplier whitelist of a leading internet enterprise, which shows it has obtained the qualification to participate in the competition, but it cannot directly prove the procurement amount.
Third, data center orders do not all equal CDU orders. The same semi-annual report disclosed that in the first half of 2026, Midea won the bid for China Unicom's centralized procurement of communication room air conditioners for over CNY 200 million, and secured a 70% share in China Mobile's centralized procurement project for low-voltage chiller units. These figures can prove its business foundation with operators, but they correspond to computer room air conditioners and chiller units, and cannot be used as CDU performance.
As of the drafting of this article, Midea has not separately disclosed the revenue, profit, backlog order amount, or customer repurchase situation of the CDU business in the aforementioned annual reports, semi-annual reports, and investor relations records.
Therefore, it can now be said that it "has been implemented," but it cannot yet be asserted based on this that it "has been realized on a large scale."
Next, what is truly worth watching is not the number of new products, but three report cards
The first is the order report card: after the Wuhu and Malaysia projects, will there be more verifiable customers, procurement amounts, and repurchase projects? A demonstration project can prove the product is usable, but continuous orders can prove customers are willing to pay repeatedly.
The second is the delivery report card: after the Shunde base is put into production, can the production line ramp up as planned, and can the products be manufactured stably and delivered on time? Since the CDU is close to the servers, stability and after-sales response will directly affect whether customers expand their procurement.
The third is the financial report card: will Midea further disclose the revenue, profitability, and order structure of the liquid cooling business? If there are always only project names without operational figures, it will be difficult for the outside world to judge exactly how big this business is.
Conclusion: What Midea Needs to Cross Is Not Just a Technical Threshold
So, why does Midea still want to make CDUs?
Because customers are pushing the cooling system from the server room to the servers, and Midea does not want to just stand at the chiller end. It hopes to continue extending its existing capabilities in cold sources, control, manufacturing, customers, and engineering delivery to the key position between the two water circuits.
This is not a sudden leap from household air conditioners into an unfamiliar track; it is more like following its familiar cooling system, step by step, moving closer to the chips.
This path has a basis and a realistic starting point: products have been launched, customers are participating in joint research, projects have been implemented, and testing and capacity are being supplemented. But how far this path can go cannot be concluded at this stage just by relying on a few buzzwords like "AI" and "liquid cooling."
CDUs are hidden in server rooms, hard for ordinary people to see. What Midea is crossing is not just the barrier between the two water circuits, but the barrier from "being able to make it" to "being able to continuously sell it."
The final answer still needs to be written by repeat orders, stable delivery, and financial data.