CITIC SEC: Huawei Connect Conference releases new AIDC energy infrastructure products, accelerating upgrades in energy storage and liquid cooling.

CITIC SEC released a research report stating that on September 17-18, Huawei held its Connect Conference, with the AIDC Infrastructure Summit held on the afternoon of the 17th, further releasing a new generation of energy and thermal management products around source-grid-load-storage AIDC, including Taishan UPS, Hengshan centralized DC UPS, the highest-density power module 5.0, intelligent lithium battery 5.0, and AI intelligent liquid cooling system. Huawei proposed a "3+1" reconstruction for AIDC covering Watt, Heat, Bit, and construction mode, and through a MIMO architecture achieved multi-power input, multi-standard output, and multi-level energy storage coordination. As AIDC develops toward hundred-MW and GW levels and per-cabinet power continues to increase, traditional data center power supply and cooling systems are accelerating their reconstruction. Among these, energy storage will upgrade from traditional backup power to load smoothing, grid friendliness, and energy regulation, while liquid cooling will also upgrade from a single cooling device to an intelligent thermal management system. Energy storage and liquid cooling are expected to become important incremental segments of AIDC infrastructure. CITIC SEC's main views are as follows: Huawei released multiple new AIDC energy infrastructure products: high-voltage DC, multi-layer energy storage, and liquid cooling have become important evolution directions. Power supply: UPS remains a core segment, and energy storage is evolving from backup power to multi-level coordination, which can also reduce the need for diesel generator configuration. Huawei this time proposed a MIMO power supply architecture, building a multi-level energy storage system from board level, cabinet/room level to campus level, respectively addressing microsecond/second-level, millisecond/minute-level, and hour-level power fluctuations, and through the coordination of UPS, intelligent lithium batteries, and campus-level grid-forming energy storage, absorbing AI load fluctuations within AIDC as much as possible. On the product side, the new-generation Taishan UPS reaches 1.28MW per cabinet with efficiency of 98%, and can smooth up to about 70% of AI load fluctuations; Hengshan centralized DC UPS is compatible with 270V, 400V, and 800V DC standards, with efficiency reaching 98.5%; intelligent lithium battery 5.0 reaches 400kW per cabinet, supports 10-minute backup power, and adapts to an 800V high-voltage DC architecture. On the campus side, grid-forming energy storage further assumes the functions of stabilizing the grid, smoothing loads, and peak shaving. Huawei proposed that through coordination between energy storage and diesel generators, diesel generator deployment can be reduced by about 30%, and during power supply peaks, energy storage can also bear about 30% of peak demand. The AIDC power supply architecture is not about energy storage replacing UPS or diesel generators, but about evolving from traditional UPS + diesel generator backup power to a multi-level coordinated system of UPS + room-level energy storage + campus-level energy storage + diesel generators; UPS continues to undertake highly reliable power supply and rapid power quality management, while energy storage adds high-frequency power regulation and energy regulation capabilities and reduces the need for redundant diesel generator configuration. As AIDC power density and campus scale increase, both the frequency of use and value content of energy storage in the power supply system are expected to rise. Liquid cooling: continuously rising per-cabinet power is driving liquid cooling to become a mandatory option for high-density AIDC, and the focus of competition is upgrading from cooling equipment to intelligent thermal management systems. With the increase in AI server power density, Huawei judges that liquid cooling has gradually become a mandatory option for AIDC and has upgraded the TMU from a traditional cooling medium distribution device to the heart + brain of the entire liquid cooling system. The AI intelligent liquid cooling system released this time covers the 450kW to 1.38MW power range, natively adapts to Ascend, and is also compatible with NVIDIA computing platforms; with the help of AI operations and maintenance models, it can identify micro-leaks of 0.02L/min and sense changes in coolant health status about 7 days in advance. In addition, the TMU can communicate directly with Ascend NPUs via electrical signals, completing cooling adjustment in as fast as about 5 seconds after computing load changes. The industrial opportunities in AIDC liquid cooling come not only from penetration rate improvement, but also from increases in power density and system value content. In the future, liquid cooling products will further upgrade from single-point equipment such as traditional cold plates and CDUs to full-link thermal management solutions composed of cold plates, piping, coolant, TMU/CDU, sensors, and control systems. The focus of industry competition will also gradually shift from cooling efficiency to full-lifecycle reliability, intelligent operations and maintenance, and computing-thermal coordination capabilities. AIDC infrastructure is evolving from traditional supporting facilities to the integration of computing power + power + energy storage + liquid cooling. Attention should be paid to incremental opportunities in energy storage and liquid cooling. This Huawei release is not a simple iteration of individual products, but a redesign around AIDC of the power supply system from the grid to the chip and the cooling system from the chip to the outdoors. As AI cluster scale continues to expand, per-cabinet power density rises, and AIDC campuses evolve toward GW level, both the technical routes and value distribution of data center infrastructure are expected to change: on the power supply side, high-voltage DC and source-grid-load-storage coordination are expected to accelerate penetration, and energy storage is upgrading from backup power equipment to important infrastructure that ensures stable AIDC operation; on the cooling side, liquid cooling is gradually becoming standard for high-density computing power and upgrading toward intelligent thermal management systems. It is recommended to continue paying attention to the actual pull from AIDC project implementation on energy storage configuration ratios, liquid cooling value content, and industry chain orders. Investment strategy: Huawei's release of new AIDC infrastructure products further validates the industrial trends in energy storage and liquid cooling. As AI server power and per-cabinet power density continue to increase, energy storage will upgrade from traditional backup power to high-power BBU and campus-level grid-forming energy storage, and liquid cooling will upgrade from a single cooling device to a full-link intelligent thermal management system. It is recommended to focus on: 1) AIDC power supply; 2) the liquid cooling segment. Risk factors: AIDC construction progress falling short of expectations; AI computing power demand growth falling short of expectations; changes in energy storage and liquid cooling technology routes; intensifying industry competition; commercialization progress of related products falling short of expectations.