UL Solutions has introduced a certification program for medium-voltage solid-state power units, targeting an emerging part of the electrical infrastructure needed to support high-density AI data centers. The program is designed to give manufacturers a defined safety-evaluation path for equipment that can convert medium-voltage AC from the grid directly to lower-voltage DC, including architectures built around 800 VDC power distribution.
The rapid expansion of artificial intelligence is creating a problem that sits outside the processors and AI models themselves: how to deliver enough electricity to increasingly power-hungry computing infrastructure.
UL Solutions is addressing one part of that challenge with a new certification program for medium-voltage solid-state power units, power-electronic systems designed to convert medium-voltage alternating current directly into the direct current used by data center equipment.
The technology is sometimes referred to as a medium-voltage solid-state transformer. Unlike a conventional transformer, which primarily changes voltage through electromagnetic induction, a solid-state system uses power electronics to manage and convert electrical power. In the architecture targeted by UL Solutions, the equipment can support direct medium-voltage-to-800 VDC conversion.
That matters because electricity may otherwise pass through multiple conversion stages before reaching servers and high-performance computing hardware. Each conversion introduces energy losses and adds equipment to the power chain. Reducing conversion steps can therefore simplify the architecture, although the actual efficiency gains depend on the implementation and the broader electrical system.
The new UL Solutions program is based on UL 2877, an Outline for Investigation for Power Supplies, Medium Voltage. The certification process evaluates areas including electrical protection, insulation, grounding, enclosures, safety interlocks, wiring and product markings. Testing also covers conditions such as elevated temperatures, excessive electrical loads, voltage spikes and short-circuit currents.
For manufacturers, the significance is less about another component certification than about establishing a defined pathway for a power architecture that is moving into AI data center deployments.
AI workloads are pushing data centers toward higher power densities. Gartner estimates worldwide data center electricity consumption will reach 565 terawatt-hours in 2026, up 26% from 2025. The research firm expects AI-optimized servers to account for 31% of data center electricity consumption this year, with their power consumption exceeding that of conventional servers in 2027.
That growth is putting power availability alongside compute, networking and cooling as a constraint on AI infrastructure expansion. Gartner’s 2026 forecast says worldwide data center power demand is expected to reach 290 gigawatts by 2030, up from 104 gigawatts in 2025.
The electrical architecture is consequently changing alongside the computing architecture. UL Solutions has previously highlighted the industry’s movement toward higher-voltage DC systems for high-density AI racks, with approximately 800 VDC systems emerging today and designs potentially moving toward 1,500 VDC. Higher voltages, however, introduce additional safety considerations that existing codes and infrastructure practices must address.
That makes certification and standards development an increasingly important layer of the AI infrastructure stack.
UL Solutions has already been working with organizations including the Open Compute Project, ABB and Eaton on safety standards for next-generation data center power distribution. That initiative includes reviewing requirements for equipment such as switchgear, panelboards, busbars, cables and overcurrent protection as AI data centers move toward higher-power architectures.
The company’s latest certification program addresses a more specific point in that evolution: the equipment that connects medium-voltage utility power with the DC architecture serving advanced computing systems.
The safety evaluation is particularly relevant because these units handle substantially more electrical power than many conventional power-supply applications. Electrical insulation, grounding, fault protection, thermal behavior and short-circuit performance become critical engineering considerations when power conversion is moved closer to high-density compute infrastructure.
The shift also reflects a broader trend in AI infrastructure: optimizing the entire path from the electric grid to the accelerator, rather than focusing exclusively on processor efficiency.
NVIDIA’s AI accelerators, for example, have driven increasingly dense compute platforms, while hyperscalers and data center operators are redesigning racks, cooling systems, networking and power distribution around those workloads. The result is an infrastructure stack in which electrical engineering increasingly becomes part of AI system scaling.
Gartner expects worldwide AI spending to reach $2.7 trillion in 2026, with AI infrastructure—including AI-optimized servers, infrastructure services, networking and processing hardware—representing a major portion of that investment.
Solid-state power conversion is therefore emerging alongside other technologies intended to make AI data centers more power-efficient and scalable. But certification remains important because greater electrical efficiency does not eliminate safety risks.
For manufacturers, UL Solutions’ program provides a framework for demonstrating compliance before this equipment reaches data center operators and code authorities. For operators, independent testing can provide another source of information when assessing unfamiliar power architectures.
The broader direction is clear: AI infrastructure is no longer just a question of supplying more GPUs. As compute densities rise, the electrical systems feeding those processors are becoming a technology category of their own.
Medium-voltage solid-state power units are one example of that transition. Their adoption will depend on efficiency, reliability, economics, interoperability and regulatory acceptance, but establishing safety requirements is an important prerequisite for deploying the technology at scale.
Market Landscape
AI data center infrastructure is increasingly being redesigned around power density and availability. Gartner projects data center electricity consumption to reach 565 TWh in 2026 and says power availability is becoming a limiting factor for AI capacity expansion.
The resulting infrastructure shift includes higher-voltage DC distribution, advanced power conversion, energy storage, on-site generation and new approaches to cooling. UL Solutions says data centers are moving toward high-voltage DC architectures to support high-density AI racks while reducing energy loss and heat generation.
Certification is becoming part of this transition. In April 2026, UL Solutions certified ABB’s HiPerGuard medium-voltage multi-mode static inverter to UL 9540, highlighting the growing role of independent testing for emerging AI data center power systems.
The new medium-voltage solid-state power-unit program extends that standards-driven approach to another emerging component of AI power infrastructure.
Top Insights
- AI data center expansion is increasing demand for power architectures capable of handling higher voltage, density and direct-current distribution.
- Medium-voltage solid-state systems can reduce conversion stages by converting grid AC directly toward DC architectures used by computing infrastructure.
- UL 2877 provides the technical foundation for evaluating electrical, thermal, insulation, grounding and fault-related safety characteristics.
- Gartner projects data center electricity consumption will reach 565 TWh globally in 2026 as AI-optimized servers drive infrastructure demand.
- Independent certification could become increasingly important as operators adopt unfamiliar power architectures for AI and high-performance computing.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
