HyT Capital Portfolio | Great Wall Power Adopts InnoScience's InnoGaN, Reconstructing AI Data Center Energy Systems
Great Wall Power has adopted InnoScience's gallium nitride (InnoGaN) technology in its titanium-class power supplies for AI data centers, achieving a power conversion efficiency of over 96% — exceeding the global highest 80PLUS Titanium energy efficiency standard.
With the growth of applications such as artificial intelligence, cloud computing, and data mining, the demand for computing power in data centers has significantly increased. According to the International Energy Agency (IEA), global data center electricity consumption will reach 3,000 TWh by 2030, accounting for approximately 10% of total global electricity consumption. By comparison, global data center electricity consumption in 2025 is estimated at around 4%.
A key factor limiting the dramatic growth of data center power consumption is power conversion efficiency. Traditional silicon-based power supplies have relatively low conversion efficiency, resulting in power losses of up to 10%. For a standard 10 MW data center, this equates to 9 million kWh of wasted electricity per year.
Every 1% improvement in power conversion efficiency can save hundreds of millions of RMB in electricity consumption. The higher efficiency of GaN-based power supplies helps eliminate this waste. At the same time, GaN also enables higher system power density, thereby reducing BOM costs and eliminating waste of BOM components ("dematerialization"). This also helps reduce greenhouse gas emissions.
In response to the energy consumption challenges of server power supplies, InnoScience has taken the lead in launching an E-GaN power IC (ISG612xTD SolidGaN) in a To-247-4 package, integrating gate drive and short-circuit protection. It supports 700V voltage and offers an Rdson range of 22~59mΩ. This series integrates a precision Vgs gate driver, features fast short-circuit protection and excellent thermal performance, and can meet the high-frequency switching requirements for Titanium Plus efficiency. Compared to conventional solutions, power density is more than doubled.
According to the OCTC *White Paper on Modular Design of High-Power-Density Server Power Supplies (2024)*, when servers are operating 80% of the time in the typical load range of 20% to 50%, GaN-based titanium-class power supplies can maintain a conversion efficiency of 95.5% to over 96%, effectively avoiding hidden energy losses caused by "efficiency gaps."
Great Wall server power supplies have taken the lead in adopting InnoScience's co-packaged chips ISG6122TD and ISG6123TD. Compared to conventional power supplies, power loss at light and medium loads is reduced by at least 30%. In the typical load range of 20% to 50%, efficiency is improved by 4 percentage points over conventional solutions, achieving a conversion efficiency of over 96%.
According to estimates, with GaN-based titanium-class power supply solutions, every 10,000 servers can save over RMB 2 million in electricity costs annually, reduce heat generation by 50%, and lower air conditioning energy consumption by 18%. This has strongly promoted the collaboration between Great Wall Power and InnoScience, accelerating the industrialization and widespread adoption of GaN titanium-class power supplies, and achieving a leap for GaN in AI data centers — from concept to application — helping to reshape the power supply system of intelligent computing centers, from component-level innovation to full energy system reconstruction.
InnoScience has already partnered with a number of data center manufacturers across the industry, developing a comprehensive portfolio of products and solutions. It is believed that in the near future, a full-link GaN data center system will be realized, contributing to energy savings and carbon reduction in data centers.
The above article is republished from the "InnoScience" 「英诺赛科」official WeChat public account.
Back
Next article
