The growth of artificial intelligence is reshaping data center design and driving a sharp rise in power consumption. Large AI facilities can draw as much electricity as a nuclear power plant, making efficiency gains at every level of the hardware stack increasingly important.
To address this issue, researchers at Jiufengshan Lab in Hubei province have unveiled a new gallium nitride-based power module. Developed under the leadership of Li Sichao and officially released in late 2025, the module replaces conventional silicon chips with gallium nitride, a third-generation semiconductor known for its higher efficiency and power density.
According to the research team, the GaN module reduces electrical power loss by around 30% while also shrinking module volume by a similar margin. At the same time, production costs are cut by roughly half, addressing both operational and economic pressures faced by data center operators.
In a 1-gigawatt AI data center, the adoption of these modules could save nearly 300 million kilowatt-hours of electricity each year, translating into an estimated CNY 240 million in annual cost savings. This is particularly relevant given that power conversion for processors accounts for about 11% of a data center’s total energy consumption. By improving efficiency in a core but often overlooked component, the Jiufengshan Lab development shows how material innovation can deliver tangible benefits for AI infrastructure. For telecom operators, cloud providers, and hyperscale data center owners, such advances may play a key role in keeping AI growth both economically and environmentally sustainable.
