SolarEdge displayed its MVAC-to-800 VDC solid-state transformer technology for AI data centers at the OCP Global Summit.
At the OCP Global Summit last week, SolarEdge put its AI data center power ambitions on display, quite literally. The company brought a working solid-state transformer cell and a 3D-printed model of its containerized data center power architecture to the NVIDIA AI Factory MGX Ecosystem Showcase, where it demonstrated its MVAC-to-800 VDC AI Factory Power Technology to an industry audience gathered October 12–15.
The OCP Global Summit is one of the industry's main forums for open compute infrastructure, the kind of venue where serious hardware gets serious attention. SolarEdge chose it deliberately. The NVIDIA AI Factory MGX Ecosystem Showcase, running October 12–15 within the summit, gave the company direct access to the data center engineering community it is targeting.
The physical exhibits told the story without much need for explanation. A working solid-state transformer cell showed the core conversion technology in hardware form, while the 3D-printed model of SolarEdge's containerized data center power architecture showed how that technology fits into a broader grid-to-rack system. An abstract development roadmap became something attendees could actually see.
The white paper collaboration with NVIDIA and the decision to show up at OCP both suggest the company is working to build credibility and relationships in the data center market early.
**Why SolarEdge is developing an 800 VDC power path for AI infrastructure**
AI infrastructure is pushing power density to levels that existing data center electrical systems were not built to handle. As GPU clusters scale up, the industry has to rethink how electricity moves from the utility grid all the way down to the server rack, through every conversion stage, every distribution run, and every protection point.
SolarEdge is approaching that problem from a specific angle. The company has spent roughly two decades building DC power electronics for photovoltaic applications, including inverters, optimizers, protection systems, and monitoring platforms. That is a deep base of DC engineering experience, and it is now being applied directly to data center requirements.
The goal is a full DC powertrain: a system covering medium-voltage AC conversion, 800 VDC distribution, protection, monitoring, and rack power input. The aim is a complete chain from grid connection to compute load.
**Current development status and collaboration with NVIDIA**
This is not just a concept. SolarEdge says the three-phase MVAC-to-800 VDC conversion stage is already running under load in its engineering labs, and system-level validation of the full power path is underway, though no commercial availability date has been announced.
That lab work runs alongside a published collaboration with NVIDIA. In September, SolarEdge released a white paper titled *800 VDC Protection and Grounding for AI Data Centers*, developed in collaboration with NVIDIA, covering grounding strategies, fault behavior, and protection coordination for 800 VDC architectures.
SolarEdge positioned the white paper as technology-neutral, a framework for multi-vendor industry evaluation rather than a blueprint for a single proprietary implementation. That framing signals the company wants to participate in an emerging ecosystem rather than lock it down.
**Technical features built into the architecture**
The architecture integrates several layers of electrical safety and monitoring: branch-level sensing and interruption, insulation monitoring, and system-level monitoring, all designed in from the start rather than added afterward.
That level of built-in protection reflects the realities of DC distribution at scale. High-voltage DC systems behave differently from AC systems when faults occur, so protection coordination has to be deliberate. SolarEdge's PV background is directly relevant here, since DC arc fault detection and insulation monitoring are standard requirements in solar installations. The design spans from medium-voltage conversion all the way to the rack, consolidating what is currently a fragmented chain of separate conversion and distribution components, consistent with the OCP community's broader philosophy around open, multi-vendor ecosystems.
**Background: SolarEdge's position in the energy and data center markets**
SolarEdge Technologies is primarily known as a solar power electronics company. Its DC-optimized inverters and power optimizers are installed in more than 146 countries, and its core business spans energy management, storage, and grid services, almost entirely in the photovoltaic sector.
The move into AI data center power infrastructure extends that foundation rather than departing from it. Managing high-voltage DC, protecting against faults, and monitoring distributed systems are engineering disciplines that transfer directly. The application is new; the expertise is not.
The data center market is, however, a different competitive environment from the solar industry. SolarEdge will be entering a space with established players and rapidly evolving standards. The white paper collaboration with NVIDIA and the decision to show up at OCP both suggest the company is working to build credibility and relationships there early.
**What SolarEdge's OCP showcase signals**
SolarEdge showed up at OCP with hardware and a published technical document, two concrete outputs from a development program still in the validation phase. The solid-state transformer cell running under load in the lab, the white paper developed with NVIDIA, and the containerized architecture model on the show floor all point to a company moving methodically from concept toward product.
The 800 VDC distribution path for AI data centers remains an emerging standard, not a settled one. SolarEdge's positioning (open ecosystem, technology-neutral frameworks, multi-vendor compatibility) reflects where the industry actually is right now: early enough that the standards are still being written, and competitive enough that being part of that conversation matters.