News: Microelectronics
8 June 2026
onsemi’s role in NVIDIA MGX ecosystem expanding into 800VDC power architectures
As hyperscalers and enterprise operators race to build increasingly powerful AI infrastructure, power delivery and energy efficiency are emerging as the most critical constraints. Industry analysts estimate that AI rack power requirements could grow beyond 1MW/rack in the very near future.
Against that backdrop, intelligent power and sensing technology firm onsemi of Scottsdale, AZ, USA is expanding its role within NVIDIA’s MGX™ ecosystem to supply power systems designed to support the next generation of AI data centers and accelerated computing platforms.
NVIDIA’s MGX architecture
The NVIDIA MGX modular building block architecture allows original equipment manufacturers (OEMs) and system builders to more rapidly configure, deploy and scale AI infrastructure while reducing development complexity and time to market. onsemi provides technology that drives every stage of power conversion within the MGX ecosystem.
onsemi’s broad portfolio across the power tree positions it to support the mix-and-match flexibility that is central to the MGX ecosystem. The firm’s internal fabrication capacity and expertise across silicon, silicon carbide and gallium nitride technologies also provide supply chain resiliency as AI infrastructure deployments accelerate globally.
onsemi technologies already support existing MGX systems, with growing opportunities across power field-effect transistors (FETs), multi-phase power solutions, silicon carbide (SiC) JFETs and gallium nitride (GaN) technologies. The collaboration is also expected to expand into emerging 800VDC power architectures designed for increasingly compute-dense AI environments.
“As explosive growth in AI adoption, agentic AI and reasoning sophistication drives exponential growth in AI computing, the demands on power delivery, cooling and infrastructure efficiency are increasing dramatically,” notes Sudhir Gopalswamy, president of onsemi’s Intelligent Sensing and Analog and Mixed-Signal Group. “High-voltage architectures such as emerging 800VDC designs are essential to scale AI infrastructure. They increase power density to 1MW/rack and beyond, with improved efficiency and reduced deployment costs and complexity.”
Solving one of AI’s biggest infrastructure bottlenecks
onsemi’s role within the MGX ecosystem centers on solving one of AI infrastructure’s most pressing bottlenecks: how to deliver more power, more efficiently, into increasingly dense compute environments.
The firm provides power solutions directly to Nvidia and the MGX ecosystem partners, ranging from power supply units (PSU) to battery backup unit (BBU) providers to 800VDC PDB suppliers that enables the modularity of the MGX system.
Power efficiency in the AI economy
Industry analysts increasingly view power delivery as one of the defining constraints of the AI era. As GPU clusters scale from individual servers to rack-scale and data-center-wide deployments, even incremental efficiency improvements can translate into meaningful reductions in operating costs and energy consumption.
onsemi says that it plays a critical role in the NVIDIA MGX ecosystem as a power semiconductor manufacturer that possesses the technical depth across technologies able to support increasingly complex architectures.
Beyond the data center
The implications extend beyond existing AI data centers. The same power architectures being developed for AI factories are expected to influence future designs across automotive and industrial applications — markets where efficiency, thermal management and system reliability remain paramount.
In that sense, the NVIDIA MGX collaboration demonstrates onsemi’s expertise and robust power portfolio that enables the foundational infrastructure layer of the AI economy itself.
And as the race to build the next generation of AI factories intensifies, power efficiency may prove just as important as processing power, says onsemi. The firm reckons that it is positioned to address these needs both from a technical as well as an in-house manufacturing standpoint.
