Huawei Unveils Advanced Power and Cooling Solutions for AI Data Centers

Huawei unveiled a new generation of energy and thermal management products for Artificial Intelligence Data Centers (AIDC) at its Huawei Connect conference on September 17-18. The announcement, detailed in a research note from CITIC Securities, highlights a strategic push to re-engineer data center infrastructure to support the growing power and cooling demands of large-scale AI clusters.

New Infrastructure for the AI Era

At its AIDC Infrastructure Summit, Huawei introduced a framework for redesigning data centers that focuses on power, heat, data, and construction models. The company is advancing a Multiple-Input Multiple-Output (MIMO) power architecture designed to coordinate multiple power sources, output formats, and layers of energy storage.

Key product releases include:

• Taishan UPS: A new uninterruptible power supply with a single-cabinet power of 1.28 MW and 98% efficiency, designed to smooth up to 70% of AI workload fluctuations.
• Hengshan Centralized DC UPS: A direct current UPS with 98.5% efficiency, compatible with 270V, ±400V, and 800V systems.
• Smart Lithium Battery 5.0: A 400 kW battery cabinet providing 10 minutes of backup power, adapted for high-voltage DC architectures.
• AI Smart Liquid Cooling System: An intelligent thermal management solution covering a power range of 450 kW to 1.38 MW.

A Shift in Power and Cooling Strategy

Ad

According to the CITIC Securities analysis, these products reflect a fundamental shift in how AIDC infrastructure is designed. Energy storage is evolving from a simple backup power source to a multi-layered system that provides load smoothing, grid stability, and energy regulation. Huawei's approach aims to create a coordinated system of UPS, room-level storage, campus-level storage, and diesel generators, which it claims could reduce generator deployment by approximately 30%.

On the cooling front, increasing server power density is making liquid cooling a necessity. Huawei's new AI-powered system moves beyond individual components to offer a full-stack thermal management solution. The system is designed to communicate directly with Huawei's Ascend processors to adjust cooling within five seconds of a load change and uses AI models to detect minute leaks of 0.02 L/min and predict coolant health issues up to seven days in advance.

Market Implications

CITIC Securities views these developments as validation of key industry trends. As AI data centers scale to the megawatt and gigawatt level, infrastructure is moving from a supporting role to an integrated system of computing, power, and thermal management. The report identifies energy storage and liquid cooling as the most significant growth areas within this new paradigm.

Analysts suggest that as AIDC projects are deployed, the market should monitor the actual adoption rates of advanced energy storage and the increasing value of integrated liquid cooling systems. The report also noted potential risks, including slower-than-expected AIDC construction, fluctuations in AI computing demand, and increased industry competition.