Huawei Unveils Grid-Interactive AIDC Solution at HUAWEI CONNECT 2026
At the recently held HUAWEI CONNECT 2026 in Shanghai, Huawei introduced its cutting-edge Grid-Interactive AIDC solution. This innovation is poised to redefine the landscape of artificial intelligence infrastructure, emphasizing the enhancement of energy utilization efficiency by maximizing token processing per watt.
The event commenced with opening remarks from Hou Jinlong, a board member of Huawei and president of Huawei Digital Power. He elaborated on four fundamental principles aimed at addressing vital challenges faced during AIDC center constructions. These challenges include ensuring adequate power supply, maintaining energy quality, cooling server cabinets, and implementing effective infrastructure.
Firstly, the new power architecture, complemented by liquid cooling solutions, guarantees stable power delivery and efficient heat dissipation, thereby boosting the overall system's reliability and effectiveness. Secondly, a multi-layered hybrid energy storage system helps balance local power fluctuations, improving power quality and interaction with the electric grid.
Thirdly, by closely coordinating computational power with energy systems, and fostering mutual cooperation between energy supply and consumption, the AIDC can transform from a passive energy consumer to an active component of a modern energy ecosystem. Lastly, a novel construction approach utilizing modular, prefabricated, and product-oriented solutions significantly reduces project turnaround times, enhancing operational efficiency.
Hou emphasized that the competitive edge of AI computing centers will hinge on their capability to maximize token processing efficiency while minimizing individual token costs. Alongside him, Bob He, Huawei's vice president for digital power, pointed out that as renewable energy sources become increasingly integrated into the power grid, energy electronic devices are gaining a central role. Concurrently, the surging demands of AI workloads are causing rapid fluctuations in power consumption, dictating the need for synchronized energy and computational infrastructures.
Huawei's Grid-Interactive AIDC solution, built on a "3+1" innovation framework, aims to deliver high reliability, energy efficiency, rapid deployment, and improved grid interaction, thereby optimizing token output per watt. The proposed architecture is also future-ready, responding to industry challenges through its Grid-Interactive AIDC 1.0 configuration. This system integrates network-compatible UPS (Uninterruptible Power Supply), intelligent lithium batteries, and energy storage units that operate in grid-forming modes, adeptly adjusting to the power grid's operational conditions while supporting its stability and addressing fluctuations arising from AI system demands.
To ensure a superior cooling method, Huawei showcased a liquid cooling system that employs AI for assessing working fluid conditions and bolstering predictive maintenance. He asserted that Huawei's advantage lies in its comprehensive skillset covering the entire energy chain, coupled with an extensive suite of AI technologies.
As AI continues to evolve rapidly, the computational demands are exponentially increasing, which places significant stress on existing AI infrastructures. Xia Qin, Huawei's chief expert for computational strategy, noted that AIDC is trending towards an integrated and coordinated framework that encompasses computing power, networking, and storage systems, with energy serving as the backbone of the entire setup.
A reliable, energy-efficient, and swiftly deployable power system is essential for the stable operation of substantial AI computing clusters, paving the way for extensive model training and inference in the age of AI agents. While conventional data centers have limited interaction with the power grid, AI workloads can fluctuate vastly and frequently, causing grid management issues such as outages and unstable broadband oscillations.
Josh Chen, founder and president of VNET Group, suggested redesigning power sources, data center campuses, and buildings for mutual energy supplementation, allowing for effective load management across campuses and transitioning buildings to direct current supply. This shift would create a new power system based on the direct utilization of green energy and active microgrids, ensuring stable and reliable power for AIDC systems with gigawatt capabilities.
Lin Hai, the general manager of the Intelligent Computing Center within SenseTime's SenseCore Business Group, discussed at the summit the evolution of measuring computational efficiency within SenseTime's systems—from the Power Usage Effectiveness (PUE) metric to a new Token Processing Watts (TPW) measurement. This new metric directly connects energy consumption with token output and covers the entire operational chain, from power plants and electric networks through data center power supplies and cloud services to AI model functionalities. According to the company, this approach more accurately reflects the real efficiency of AI processes, leading SenseCore to boost its TPW by 80% through comprehensive optimizations across the chain.
At HUAWEI CONNECT 2026, Huawei not only showcased the Grid-Interactive AIDC solution but also a comprehensive AI inference solution designed to provide reliable, energy-efficient, and rapidly deployable infrastructures for AI clusters.
The event commenced with opening remarks from Hou Jinlong, a board member of Huawei and president of Huawei Digital Power. He elaborated on four fundamental principles aimed at addressing vital challenges faced during AIDC center constructions. These challenges include ensuring adequate power supply, maintaining energy quality, cooling server cabinets, and implementing effective infrastructure.
Firstly, the new power architecture, complemented by liquid cooling solutions, guarantees stable power delivery and efficient heat dissipation, thereby boosting the overall system's reliability and effectiveness. Secondly, a multi-layered hybrid energy storage system helps balance local power fluctuations, improving power quality and interaction with the electric grid.
Thirdly, by closely coordinating computational power with energy systems, and fostering mutual cooperation between energy supply and consumption, the AIDC can transform from a passive energy consumer to an active component of a modern energy ecosystem. Lastly, a novel construction approach utilizing modular, prefabricated, and product-oriented solutions significantly reduces project turnaround times, enhancing operational efficiency.
Hou emphasized that the competitive edge of AI computing centers will hinge on their capability to maximize token processing efficiency while minimizing individual token costs. Alongside him, Bob He, Huawei's vice president for digital power, pointed out that as renewable energy sources become increasingly integrated into the power grid, energy electronic devices are gaining a central role. Concurrently, the surging demands of AI workloads are causing rapid fluctuations in power consumption, dictating the need for synchronized energy and computational infrastructures.
Huawei's Grid-Interactive AIDC solution, built on a "3+1" innovation framework, aims to deliver high reliability, energy efficiency, rapid deployment, and improved grid interaction, thereby optimizing token output per watt. The proposed architecture is also future-ready, responding to industry challenges through its Grid-Interactive AIDC 1.0 configuration. This system integrates network-compatible UPS (Uninterruptible Power Supply), intelligent lithium batteries, and energy storage units that operate in grid-forming modes, adeptly adjusting to the power grid's operational conditions while supporting its stability and addressing fluctuations arising from AI system demands.
To ensure a superior cooling method, Huawei showcased a liquid cooling system that employs AI for assessing working fluid conditions and bolstering predictive maintenance. He asserted that Huawei's advantage lies in its comprehensive skillset covering the entire energy chain, coupled with an extensive suite of AI technologies.
As AI continues to evolve rapidly, the computational demands are exponentially increasing, which places significant stress on existing AI infrastructures. Xia Qin, Huawei's chief expert for computational strategy, noted that AIDC is trending towards an integrated and coordinated framework that encompasses computing power, networking, and storage systems, with energy serving as the backbone of the entire setup.
A reliable, energy-efficient, and swiftly deployable power system is essential for the stable operation of substantial AI computing clusters, paving the way for extensive model training and inference in the age of AI agents. While conventional data centers have limited interaction with the power grid, AI workloads can fluctuate vastly and frequently, causing grid management issues such as outages and unstable broadband oscillations.
Josh Chen, founder and president of VNET Group, suggested redesigning power sources, data center campuses, and buildings for mutual energy supplementation, allowing for effective load management across campuses and transitioning buildings to direct current supply. This shift would create a new power system based on the direct utilization of green energy and active microgrids, ensuring stable and reliable power for AIDC systems with gigawatt capabilities.
Lin Hai, the general manager of the Intelligent Computing Center within SenseTime's SenseCore Business Group, discussed at the summit the evolution of measuring computational efficiency within SenseTime's systems—from the Power Usage Effectiveness (PUE) metric to a new Token Processing Watts (TPW) measurement. This new metric directly connects energy consumption with token output and covers the entire operational chain, from power plants and electric networks through data center power supplies and cloud services to AI model functionalities. According to the company, this approach more accurately reflects the real efficiency of AI processes, leading SenseCore to boost its TPW by 80% through comprehensive optimizations across the chain.
At HUAWEI CONNECT 2026, Huawei not only showcased the Grid-Interactive AIDC solution but also a comprehensive AI inference solution designed to provide reliable, energy-efficient, and rapidly deployable infrastructures for AI clusters.
Topics Consumer Technology）
【About Using Articles】
You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.
【About Links】
Links are free to use.