#China #Shanghai #Huawei #AI Infrastructure #AIDC

Huawei Connect 2026: Transforming AI Infrastructure with AIDC Solutions

On September 21, 2026, Huawei opened the doors of Huawei Connect 2026 in Shanghai, where a pivotal event took place—the Huawei AIDC Facility Summit. At this summit, the company introduced its interactive AIDC solutions designed to revolutionize AI infrastructure, focusing on enhancing efficiency and maximizing energy utilization.

New Paradigm in AI Infrastructure

Hou Jinlong, Huawei's Board Member and President of Digital Power, delivered the opening speech, outlining the architectural innovations necessary to tackle challenges in AIDC systems. He emphasized four key strategies:

1. Power Architecture and Liquid Cooling: These innovations ensure a stable power supply while enhancing heat dissipation, thereby maximizing reliability and efficiency.

2. Multi-layer Energy Storage System: This system absorbs local power fluctuations, which not only enhances AIDC power quality but also ensures compatibility with the electrical grid.

3. Synergy Between Power and Computing: The interaction between electric supply and consumption enables AIDC systems to transition from rigid loads to integral components of a modern electricity framework.

4. Modular Construction and Delivery: Innovations in construction methods significantly shorten project delivery times, enhancing the supply chain.

Maximizing the number of tokens per watt while minimizing token costs will be crucial for the competitiveness of computing factories, according to Hou.

Embracing Renewable Energy

Bob He, Vice President of Huawei Digital Power, highlighted the increasing role of renewable energy in the electrical grid. He noted that AI workloads exhibit rapid and sharp power fluctuations, underscoring the need for a tightly integrated computing and energy infrastructure. To address this, Huawei developed a reliable, energy-efficient, and quickly implementable AIDC solution compatible with the grid, aiming to maximize tokens per watt.

The MIMO Power Supply Architecture proposed by Huawei meets the challenges of the sector. The rollout of AIDC 1.0, which is grid-compatible, integrates a grid-supported UPS, smart lithium batteries, and energy storage systems, allowing AIDC to adapt to and support the electrical grid while smoothing out fluctuations in AI workloads.

In refrigeration, Huawei unveiled an AI-driven liquid cooling system that anticipates the state of working fluids, aiding predictive maintenance.

Establishing Foundations for the AI Era

The rise of AI demands higher productivity from infrastructure. As per Xia Qin, Chief Expert in Strategic Planning at Huawei, AIDC is moving toward an integrated architecture that combines computing, networking, and storage. Reliable, energy-efficient, and quickly deployable energy systems are essential for the seamless operation of large-scale AI computing clusters, facilitating extensive training and inference by AI agents.

Traditional data centers often interact minimally with the power grid, but with AI workloads prone to significant fluctuations, this limited interaction can lead to network failures or unstable bandwidth oscillations. Josh Chen, Founder and President of VNET Group, discussed redefining energy sources, facilities, and buildings to promote mutual support among energy sources, enhancing supply chain planning.

Transitioning to Energy Efficiency

The summit addressed the evolution from Power Usage Effectiveness (PUE) to Tokens Per Watt (TPW) in the sector. Lin Hai, General Manager at SenseTime’s Intelligent Computing Center, explained how this shift links energy consumption directly with token production across the entire energy value chain, which is crucial for measuring AI performance. By optimizing from end to end, SenseCore has improved TPW by 80%.

During Huawei Connect 2026, the introduction of the interactive AIDC solution alongside comprehensive AI inference solutions demonstrates Huawei's commitment to building a highly reliable, energy-efficient, and rapidly deployable computing infrastructure for AI clusters. This new paradigm marks a significant step toward transforming the landscape of AI infrastructure globally, enabling enterprises to harness computational power effectively.