AES Semigas

Honeywell

1 August 2025

Innoscience collaborating with NVIDIA on 800VDC power architecture for AI data centers

InnoScience (Suzhou) Technology Holding Co Ltd — which manufactures gallium nitride on silicon (GaN) power chips on 8” silicon wafers — says that, with large AI models and high-performance computing placing higher demands on power supply systems, it is collaborating with NVIDIA of Santa Clara, CA, USA to jointly promote the large-scale implementation of an 800VDC power architecture in AI data centers.

NVIDIA’s 800VDC architecture is the latest generation of power systems specifically designed to efficiently power future megawatt-scale computing infrastructure. Compared with traditional 54V power systems, the 800VDC architecture offers significant advantages in system efficiency, heat dissipation and reliability, enabling it to support a 100x to 1000x increase in AI computing power.

Innoscience says that its third-generation GaN devices offer exceptional high frequency, high efficiency and high power density. They provide NVIDIA’s 800VDC architecture with a full-link GaN power supply solution, from 800V input to GPU terminals, spanning 15V to 1200V. With the deep integration of 800VDC power architecture and GaN technology, AI data centers can achieve a quantum leap from kilowatts to megawatts in the coming years, it is reckoned, ushering in an era of more efficient, reliable and greener AI computing.

See related items:

Innoscience and Midea partner to accelerate GaN adoption in home appliance industry

InnoScience floats in IPO on Hong Kong Stock Exchange

Tags: GaN-on-Si

Visit: www.innoscience.com

RSS

Microelectronics UK

Book This Space