Huawei Grid-Interactive AIDC Solution: Shaping the New Paradigm of AI Infrastructure

SHANGHAI, Sept. 20, 2026 /PRNewswire/ -- HUAWEI CONNECT 2026 kicked off in Shanghai, The HUAWEI AIDC Facility Summit was held alongside the event. Huawei unveiled its grid-interactive AIDC solution, setting a new paradigm for AI infrastructure to maximize tokens per watt.

Hou Jinlong, Director of the Board of Huawei and President of Huawei Digital Power, delivered the opening speech at the summit. He shared four insights on architectural innovation and systematic design to address AIDC construction challenges such as power acquisition, power quality, rack power supply and heat dissipation, and engineering delivery.

First, innovative power architecture and liquid cooling ensure stable power supply and efficient heat dissipation, delivering comprehensive reliability and efficiency. Second, a multi-layer hybrid energy storage system absorbs local power fluctuations, enhancing AIDC power quality and grid-friendliness. Third, electricity-compute synergy and supply-consumption interaction enable AIDCs to transform from rigid loads to an integral part of the new power system. Fourth, construction mode innovation and modular, prefabricated, and product-based delivery significantly shorten the lead time.

Mr. Hou emphasized that maximizing tokens per watt and minimizing costs per token will be the core competitiveness of compute factories.

Huawei Grid-Interactive AIDC Solution Unveiled

Bob He, Vice President of Huawei Digital Power, stated that as the share of renewable energy grows, the power grid exhibits stronger features of power electronics. Coupled with the rapid and severe power fluctuations of AI workloads, compute and energy infrastructure are no longer decoupled but deeply coordinated. By leveraging "3+1" innovations, Huawei builds a grid-interactive AIDC solution that is highly reliable, energy-efficient, fast to deliver, and grid-friendly to maximize tokens per watt.

In the watt domain, Huawei proposes a future-oriented MIMO power supply architecture. To address industry challenges, Huawei has launched the grid-interactive AIDC 1.0 solution. By integrating grid-friendly UPS, intelligent lithium batteries, and grid-forming energy storage systems, the solution enables AIDC to adapt to and support the power grid while smoothing AI workload fluctuations. In the heat field, Huawei launched an AI-powered liquid cooling system, predict the health of working fluids, and facilitate predictive maintenance.

According to Mr. He, Huawei's unique advantage lies in its end-to-end energy and full-stack AI capabilities.

Compute-Electricity Synergy: Establishing a Solid Foundation for the AI Era

Booming AI has brought about a sharp increase in tokens, posing higher requirements on AI infrastructure. Xia Qin, Chief Expert in Huawei Computing Strategy Planning, noted that AIDC is evolving toward an integrated, coordinated compute-network-storage architecture, and energy is fundamental to infrastructure. An energy system that is highly reliable, energy-efficient, and rapidly deployable allows hyperscale AI compute clusters to operate reliably, establishing a solid foundation for large-scale training and inference in the agentic era.

Traditional data centers have limited interaction with the power grid. However, AI workloads are prone to high-frequency changes, which can easily cause grid operation issues such as grid disconnection and unstable wideband oscillation. Josh Chen, Founder & Chairman of VNET Group, stated that power sources, campuses, and buildings can be redefined to enable mutual support among power sources, scheduling within campuses, and DC power supply for buildings. This will create a new power system centered on direct green power connection and active microgrids, ensuring reliable and stable power supply for GW-level AIDC.

Lin Hai, General Manager of Intelligent Computing Center, SenseCore Business Group at SenseTime, elaborated on the efficiency evolution of SenseTime intelligent computing from PUE to TPW at the summit. SenseCore is driving the transformation of IDC from energy efficiency management centered on power usage effectiveness (PUE) to end-to-end efficiency management centered on tokens per watt, which directly connects power consumption with token output, covering the entire chain from power plants and grids to IDC rack services, computing cloud services, and model services. It's a metric that reflects the essence of AI production. Through end-to-end optimization, SenseCore has improved the TPW by 80%.

At HUAWEI CONNECT 2026, Huawei showcased its grid-interactive AIDC solution and one-stop AI inference solution, to build a computing foundation that is highly reliable, energy-efficient, and fast to deliver for AI cluster infrastructure.