#China #Shanghai #Huawei #AI Infrastructure #AIDC

Huawei's Cutting-Edge AIDC Solution Unveiled at HUAWEI CONNECT 2026

On September 20, 2026, the HUAWEI CONNECT 2026 conference commenced in Shanghai, presenting a significant leap in artificial intelligence infrastructure. During the summit, Huawei unveiled its innovative AIDC (Artificial Intelligence Data Center) solution, which focuses on enhancing interactions with power grids. This groundbreaking initiative aims to establish a new model for AI infrastructure, optimizing energy usage and maximizing tokens per watt.

Key Insights from Hou Jinlong

Hou Jinlong, a board director of Huawei and president of Huawei Digital Power, delivered an insightful keynote at the summit. He emphasized four major considerations revolving around architectural innovations and systemic design to address challenges encountered in AIDC deployment, including energy procurement, power quality, panel power supply, and effective heat dissipation.

1. Innovative Power Architecture and Cooling: The use of advanced power architecture alongside liquid cooling guarantees a consistent energy supply while efficiently managing heat, ensuring reliability and efficiency.

2. Multi-layered Hybrid Energy Storage Systems: These systems mitigate local voltage fluctuations, enhancing energy quality and fostering sustainability within the power grid.

3. Synergy Between Electricity and Computing Power: The integration of electricity supply and energy consumption enables AIDC to evolve from rigid loads to essential components of the new energy ecosystem.

4. Revolutionizing Construction Methods: The shift towards modular and prefabricated, product-oriented delivery significantly reduces implementation timeframes.

Maximizing Energy Efficiency

Mr. Jinlong underscored that the effective maximization of tokens per watt and reduction of costs per token would be crucial competitive advantages for data centers. In line with the growing share of renewable energy resources within power grids, Bob He, vice president of Huawei Digital Power, explains that the energy infrastructure must adapt to the rapid and substantial power variations typical in AI workloads. The advanced solution from Huawei aims to tightly integrate computing and energy infrastructures, allowing for efficient management of energy resources.

Following these innovations, Huawei has introduced the AIDC 1.0 solution, designed to seamlessly interact with power grids. This cutting-edge solution utilizes grid-friendly uninterruptible power supplies (UPS), intelligent lithium batteries, and energy storage systems to stabilize load fluctuations created by AI tasks.

Moreover, Huawei has also unveiled an AI-powered liquid cooling system to facilitate predictive maintenance, ensuring longevity and reliability of the AIDC operations. The strength of Huawei lies in its comprehensive energy solutions and AI capabilities that span the entire data storage landscape.

The Future of AI Infrastructure

As AI technology progresses at an accelerated pace, the demand for robust AIDC infrastructure increases concurrently. Xia Qin, Huawei’s chief expert in computational technology strategy, noted that AIDC is evolving towards an integrated architecture encompassing the interplay of computational capacity, networking, and energy storage; energy is critical for this infrastructure.

Huawei's high-reliability, energy-efficient, and rapidly deployable energy systems are pivotal in supporting expansive AI computational clusters, thereby laying a solid foundation for vast training and inference operations.

Traditional data centers lack sufficient interaction with electrical power grids. Given the drastic task fluctuations entailed in AI operations, networks may face disruptions, potentially leading to disconnections and broader operational instability. Josh Chen, founder and chairman of VNET, asserted that resources, sites, and buildings could be redefined to foster mutual support between energy sources, campuses, and direct power feeding to facilities, facilitating a new energy system focused on direct green energy connections and active microgrids.

Transforming AI Network Operations

In a detailed presentation, Lin Hai, General Manager of the Intelligent Computing Center in the SenseCore Business Group, highlighted the evolution of data center efficiency from the PUE (Power Usage Effectiveness) metric to TPW (Tokens per Watt). SenseCore advocates for transforming data center management from energy efficiency focusing on PUE to a comprehensive efficiency guided by token output, covering the entire energy generation to AI service delivery chain. This metric reflects AI production essentials, and through sophisticated optimization, SenseCore has achieved an 80% improvement in TPW.

In conclusion, Huawei’s launch of its AIDC solution at the HUAWEI CONNECT 2026 aims to construct a highly reliable, energy-efficient, and rapidly deliverable computing foundation for AI cluster infrastructures, paving the way for the robust future of AI and energy integration.