--- title: "Behind the Expansion of the NVIDIA MGX Ecosystem: An Efficiency Revolution from 800V to GPU Cores Is Quietly Unfolding" type: "News" locale: "en" url: "https://longbridge.com/en/news/288070919.md" description: "Within the NVIDIA MGX ecosystem, INNOSCIENCE is advancing all-gallium nitride (All-GaN) power conversion technology, achieving full-link coverage from 800V high-voltage distribution to GPU core power supply. This technology not only breaks through the upper limit of rack power density, raising full-load efficiency of front-end conversion to 98.2%, but also significantly reduces cooling requirements and operating costs for AI factories by introducing a vertical power delivery architecture" datetime: "2026-05-29T13:08:05.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/288070919.md) - [en](https://longbridge.com/en/news/288070919.md) - [zh-HK](https://longbridge.com/zh-HK/news/288070919.md) --- # Behind the Expansion of the NVIDIA MGX Ecosystem: An Efficiency Revolution from 800V to GPU Cores Is Quietly Unfolding As artificial intelligence workloads expand toward rack-level systems and full data center scales, power delivery capability has become a core bottleneck constraining data center system performance, density, and total cost of ownership (TCO). Within the ecosystem of NVIDIA MGX, an open modular reference architecture, an efficiency revolution supported by all-gallium nitride (All-GaN) technology is quietly reshaping the power delivery path from high-voltage distribution all the way to GPU cores. The latest development in this technological evolution comes from INNOSCIENCE, a member of the NVIDIA MGX ecosystem. The company is promoting All-GaN power conversion technology covering the entire chain to support next-generation high-density AI systems. For investors and data center operators, this upgrade in underlying power semiconductor technology is crucial for breaking through the upper limits of rack power density and substantially reducing the operating costs of high-compute facilities. Traditional power delivery models are showing strain in coping with ever-increasing rack power. The challenge is no longer just bringing electricity into the rack, but how to efficiently and compactly convert high-voltage electricity into the operating voltage required by GPUs. **GaN technology, with its characteristics of low on-resistance, low gate charge, and zero reverse recovery, is becoming the 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 these power solutions that break through physical space and thermodynamic limitations. This will not only shorten the engineering R&D cycle for accelerated computing systems but also greatly accelerate the large-scale commercial deployment of next-generation AI factories. ## Front-End Conversion Breakthrough: Peak Efficiency of 12kW Solution Approaches 99% As AI rack power continues to climb, the front-end conversion stage has become one of the most demanding links in the power architecture. In NVIDIA's 800 VDC power architecture, the number of conversion stages is reduced by delivering direct current closer to the rack. However, this requires the front end to simultaneously handle high input voltage, high conversion ratios, and constrained thermal budgets and motherboard space. INNOSCIENCE's latest data demonstrates the direct benefits of GaN in this segment. **In its 12 kW 800 V to 48 V stage design, the primary side uses 650 V GaN double-sided cooling (DSC) devices, and the secondary side uses 100 V GaN devices, achieving approximately 99% peak efficiency and 98.2% full-load efficiency at an operating frequency of 1 MHz.** Furthermore, newly released 150 V GaN devices further simplify the secondary side design, reducing the number of required synchronous rectification devices by 50%. This reduction in footprint brought about by high-frequency operation has direct commercial value for AI systems pursuing higher rack density. Beyond 48 V front-end conversion, meeting different system design requirements for motherboard space and thermal budgets requires extreme flexibility in power architecture choices. **INNOSCIENCE has expanded its All-GaN solution to cover a full range of intermediate bus voltage options from 800 V to 48 V, 12 V, and 6 V.** For 800 V to 12 V conversion, the market can now utilize 40 V GaN devices to achieve efficient synchronous rectification and improve thermal performance; for 800 V to 6 V conversion, 15 V GaN devices serve as a synchronous rectification solution, supporting lower intermediate bus architectures and thereby 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 multi-phase buck conversion. At the scale of AI factories, even minor efficiency improvements mean significant reductions in cooling requirements and operating costs. ## Vertical Power Delivery Reshapes Core Response In the final conversion stage closest to the computing core, traditional lateral power delivery faces severe challenges due to distribution losses and motherboard wiring complexity, given the extremely high current demands and the critical importance of transient response. Vertical Power Delivery (VPD) is emerging as a viable architecture that provides shorter current paths, lower parasitic losses, and higher current density. **To meet the requirements of rapid dynamic transients in GPUs, INNOSCIENCE has verified the feasibility of operating 15 V GaN HEMTs at frequencies between 3 MHz and 5 MHz, which can significantly reduce the size of required magnetic components and capacitors.** Currently, the company is developing DrGaN solutions that significantly increase bandwidth by supporting high switching frequencies, thereby reducing reliance on traditional large-capacity output capacitors. As future MGX AI systems continue to increase accelerator current density, power stages supporting VPD will become important building blocks for near-core GPU power delivery. To accelerate customer adoption cycles, INNOSCIENCE offers a series of evaluation boards and reference designs to help system designers verify GaN performance throughout the AI power tree. These platforms include a 12 kW 800 V to 48 V demonstration board, a 48 V to 12 V 4-phase GaN evaluation board, and a 6 V DrGaN evaluation board for future vertical power delivery architectures. The NVIDIA MGX ecosystem is driving the deployment of modular and scalable AI infrastructure. Against the backdrop of AI infrastructure increasingly being constrained by power, the evolution of power semiconductors must keep pace with increases in compute density. By providing comprehensive coverage from 800 VDC all the way down to GPU core voltage, higher-efficiency, higher-density AI power infrastructure is accelerating from concept to reality. ### Related Stocks - [NVDA.US](https://longbridge.com/en/quote/NVDA.US.md) - [02577.HK](https://longbridge.com/en/quote/02577.HK.md) - [07788.HK](https://longbridge.com/en/quote/07788.HK.md) - [NVDY.US](https://longbridge.com/en/quote/NVDY.US.md) - [SOXX.US](https://longbridge.com/en/quote/SOXX.US.md) - [NVDX.US](https://longbridge.com/en/quote/NVDX.US.md) - [XSD.US](https://longbridge.com/en/quote/XSD.US.md) - [XLK.US](https://longbridge.com/en/quote/XLK.US.md) - [159516.CN](https://longbridge.com/en/quote/159516.CN.md) - [512480.CN](https://longbridge.com/en/quote/512480.CN.md) - [NVDL.US](https://longbridge.com/en/quote/NVDL.US.md) - [07388.HK](https://longbridge.com/en/quote/07388.HK.md) - [SMH.US](https://longbridge.com/en/quote/SMH.US.md) - [588170.CN](https://longbridge.com/en/quote/588170.CN.md) - [SOXL.US](https://longbridge.com/en/quote/SOXL.US.md) - [NVDU.US](https://longbridge.com/en/quote/NVDU.US.md) - [512760.CN](https://longbridge.com/en/quote/512760.CN.md) - [NVDQ.US](https://longbridge.com/en/quote/NVDQ.US.md) - [NVDD.US](https://longbridge.com/en/quote/NVDD.US.md) - [588780.CN](https://longbridge.com/en/quote/588780.CN.md) - [NVDS.US](https://longbridge.com/en/quote/NVDS.US.md) - [NVD.US](https://longbridge.com/en/quote/NVD.US.md) ## Related News & Research - [Boost Run Wins $471.7M, 36-Month GPU Infrastructure Contract With Thinking Machines Lab](https://longbridge.com/en/news/287975411.md) - [Nvidia faces investor capital test as $3.7T AI IPO wave nears](https://longbridge.com/en/news/287688351.md) - [21:54 ETAIC presenteert AI-infrastructuur van de volgende generatie en organiseert strategisch panel met NVIDIA en VAST Data op COMPUTEX 2026](https://longbridge.com/en/news/287700974.md) - [PATEO CONNECT Signs Strategic Cooperation Framework with NVIDIA on In-Vehicle AI and Autonomous Driving](https://longbridge.com/en/news/287994484.md) - [Lisuan GPU sells 30,000 units despite lagging performance](https://longbridge.com/en/news/287803555.md)