NVIDIA MGX Ecosystem Expansion: A Quiet Efficiency Revolution Unfolding from 800V to GPU Cores

```

As AI workloads expand to rack-level systems and full data center scales, power delivery capability has become the core bottleneck restricting system performance, density, and total cost of ownership in data centers. Within the NVIDIA MGX open modular reference architecture ecosystem, an efficiency revolution powered by All-GaN (All-Gallium Nitride) technology is quietly reshaping the power delivery path from high-voltage distribution all the way to the GPU core.

The latest developments in this technological evolution come from Innoscience, a member of the NVIDIA MGX ecosystem. The company is advancing an All-GaN power conversion technology covering the entire chain to support next-generation high-density AI systems. For investors and data center operators, this fundamental power semiconductor upgrade is tied to breakthroughs in rack power density limits and substantial reductions in operating costs for high-performance computing facilities.

The traditional power delivery model is showing fatigue in the face of ever-increasing rack power, with the challenge no longer being just bringing power into the rack, but how to efficiently and compactly convert high-voltage electricity into the working voltage required by GPUs. GaN technology, with its features such as low on-resistance, low gate charge, and zero reverse recovery, is emerging as a key enabling technology to address this challenge, directly resulting in smaller magnetic components, better thermal performance, and lower total cost of ownership (TCO).

As AI systems move toward higher-density power architectures, the market is closely watching this power delivery solution that breaks through physical and thermodynamic limitations. Not only will this shorten the engineering R&D cycles for accelerated computing systems, but it will also greatly speed up the large-scale commercial deployment of next-generation AI factories.

Breakthrough in Front-End Conversion: 12kW Solution Achieves Peak Efficiency Close to 99%

With AI rack power continuing to climb, the front-end conversion stage is becoming one of the most demanding parts of the power architecture.

In NVIDIA’s 800 VDC power architecture, by delivering DC power directly to positions closer to the rack, the conversion stages are reduced. However, this requires the front end to handle high input voltage, high conversion ratios, and constrained thermal budgets and motherboard space.

Innoscience’s latest data demonstrates GaN’s direct benefits in this stage. In its 12 kW 800 V to 48 V design, 650 V GaN double-sided cooling (DSC) devices are used on the primary side and 100 V GaN devices on the secondary side, achieving approximately 99% peak efficiency and 98.2% full-load efficiency at 1 MHz operating frequency. Moreover, newly released 150 V GaN devices further simplify secondary side design, reducing the number of required synchronous rectifiers by 50%. This footprint reduction driven by high-frequency operation has direct commercial value for AI systems seeking higher rack density.

Beyond 48 V front-end conversion, to accommodate various design needs regarding motherboard space and thermal budgets, the choice of power architecture must be highly flexible. Innoscience has extended its All-GaN solution to cover the full range of intermediate bus voltages from 800 V to 48 V, 12 V, and 6 V.

For 800 V to 12 V conversion, the market can now use 40 V GaN devices for efficient synchronous rectification and improved thermal performance; for 800 V to 6 V, 15 V GaN devices as synchronous rectifiers enable support for lower intermediate bus architectures, simplifying the final conversion to GPU core voltage. At the critical 48 V to 12 V intermediate bus stage, Innoscience’s 100 V GaN solution optimizes multiphase buck conversion. At scale in AI factories, even small efficiency improvements mean significant reductions in cooling needs and operating costs.

Vertical Power Delivery Reshapes Core Response

In the very final stage, closest to the compute core, where current demands are extremely high and transient response is critical, conventional lateral power delivery faces severe challenges due to distribution losses and motherboard wiring complexity. Vertical power delivery (VPD) is emerging as a feasible architecture offering shorter current paths, lower parasitic losses, and higher current density.

To meet the fast dynamic transient requirements of GPUs, Innoscience has validated the feasibility of 15 V GaN HEMTs operating at 3 MHz to 5 MHz, which greatly reduces the size of required magnetic components and capacitors. The company is currently developing DrGaN solutions, which, by supporting high switching frequencies, significantly increase bandwidth and reduce reliance on traditional large output capacitors. As future MGX AI systems continue to increase the current density of accelerators, power stages supporting VPD will become the key foundation for near-core power delivery to GPUs.

To accelerate customer adoption cycles, Innoscience offers a range of evaluation boards and reference designs to help system designers validate GaN performance throughout the AI power tree. These platforms include a 12 kW 800 V to 48 V demo board, a 4-phase 48 V to 12 V GaN evaluation board, and a 6 V DrGaN evaluation board for the forthcoming vertical power delivery architecture.

The NVIDIA MGX ecosystem is driving the deployment of modular and scalable AI infrastructure. As AI infrastructure is increasingly constrained by power, the evolution of power semiconductors must keep pace with advances in compute density. By providing full coverage from 800 VDC all the way down to GPU core voltage, higher-efficiency and higher-density AI power infrastructure is accelerating from concept to reality.

Risk Warning and DisclaimerThe market involves risk; investment requires caution. This article does not constitute individual investment advice, nor does it consider the particular objectives, financial situation, or needs of any specific user. Users should consider whether any opinion, viewpoint, or conclusion in this article suits their specific circumstances. Investment based on this content is at your own risk. ```