Huawei Unveils Grid Tied AIDC Strategy to Support Massive Artificial Intelligence Energy Demand Through Integrated 4T Technology and the New TokEnergy Index Metric
At the 2026 Global AIDC Industry Summit in Dongguan, close to 1,000 leaders in the industry, as well as technicians and network carriers, gathered to explore the increasing energy demand for artificial intelligence. Centered on powering the AI era, participants explored the structural evolution of AI data centers. During the event, Huawei Digital Power unveiled its grid tied AIDC strategy and presented a Unified framework connecting modern power grids and computing infrastructure, for efficiently coping with the volatility of Large Language Model loads.
According to Hou Jinlong, President of Huawei Digital Power in a keynote speech, long term AI development is fundamentally restricted by energy supply. To cope with this issue, Huawei is combining the power, bit, heat and battery domains 4T Technology. Hou said that a full scale implementation of fault tolerant liquid cooling and end to end intelligent operation should be adapted into hyper high density compute environments, ensuring power grid friendliness and stable output at intensive compute workloads.
Huawei's new strategy is based on a 3+1 innovation model to transform electricity into computational capabilities maximally. Bob He, Vice President of Huawei Digital Power, described the technical avenues for the model as: Watt Innovation redefines power delivery from grid to chip with MIPO architecture; grid friendly UPS and grid forming ES become integrated energy router with solid state transformer (SST) eventually. Heat Innovation transcends mere equipment cooling and transforms thermal management into a whole lifecycle system. Huawei's megawatt level liquid cooling system based on AI optimizes heat dissipation in high density scenarios. Bit Innovation digital and intelligent operation of full lifecycle provide end to end visibility and automated maintenance. Construction Innovation modularization and prefabrication streamline data center building, cutting site deployment time and improving efficiency.
While efficiency is evaluated using power usage effectiveness (PUE), the increasingly dense nature of AI processing requires better metrics. To solve this, Huawei proposes the TokEnergy Index (energy/token ratio), evaluating the whole end to end efficiency to turn raw power into completed computational tokens. Thus the industry goal shifted from simply reducing power to actively creating value.
In response to this trend, Zhu Zhaosheng, Director of Huawei Computing Strategy, discussed the SuperPoD, which is built for hyperscale training models, and provides aggregated computing resources to deal with the throughput and latency issues, while requiring specifically designed power and thermal infrastructure for rapid and instant fluctuations in power load per rack. According to Teng Jianjun, Chief Architect of Huawei Cloud, modern data centers should evolve from storage space into digital factories geared for token production and be based on standardization and automation for industry growth.
