Rising AI Data Center Power Demand Drives Solid-State Transformer Commercialization, 2026 Seen as Tipping

The power supply and distribution architecture of computing centers is undergoing a transformation, with solid-state transformers (SSTs) becoming a focal point for the industry. Dozens of companies in power electronics, transmission, and distribution equipment have already entered this field. Sungrow Power Supply has released its fully self-developed solid-state transformer EnerNeo, which supports flexible combinations from 1.5 MW to 4.5 MW, adapting to different data center architectures and power granularity requirements. The product adopts a 6+3 redundancy architecture, doubling the mean time between failures.

Envision Group unveiled its AI power system in late June, featuring a core 2.5 MW solid-state transformer with a conversion efficiency of 98.5%, potentially reducing cable and copper usage by up to 80%. China XD Group, leveraging its expertise in ultra-high voltage converter valves, has been an early mover in SST research and industrialization. As the lead entity, it is spearheading the development of industry standards for solid-state transformers in data centers. The company has achieved small-scale commercial sales of its SST products and is advancing overseas delivery orders. Additionally, NARI Technology, WeiGuang Energy, Zhiguang Electric, and others have released SST products or updated related business progress.

Traditional transformers and solid-state transformers differ fundamentally in principle and function. Traditional transformers operate on electromagnetic induction, relying on line-frequency iron cores and copper windings as passive voltage conversion devices. Solid-state transformers use third-generation semiconductor devices such as silicon carbide to achieve high-frequency operation, placing them in the power electronics domain as more efficient, compact, and agile integrated power supply and distribution products. While traditional transformers are bulky, SSTs integrate silicon carbide modules to achieve a 40% to 60% reduction in volume. Conversion efficiency for traditional transformers typically ranges from 85% to 95%, whereas SSTs can achieve up to 98.5% efficiency, significantly reducing energy consumption. SSTs also enable intelligent management and digital energy management.

The 800 V DC solution based on solid-state transformers is positioned as the future technical direction for AI data center power supply and distribution. In a white paper released at the end of 2025, Nvidia explicitly identified the SST as the "ultimate goal" for next-generation AI data center power architecture, with a technology evolution roadmap: line-frequency transformer plus AC uninterruptible power supply system, 800 V high-voltage DC power supply, Panama power supply, and solid-state transformer. SSTs can be deployed locally within the computing center machine room, supporting modular maintenance, thus saving space and investment costs. They also adapt to the 800 V DC power supply architecture, eliminating low-voltage distribution cabinets, centralized UPS systems, and multi-stage rectification, improving system efficiency by 3% to 5% while reducing electricity costs and cooling energy consumption.

Key technical challenges for solid-state transformers include medium-voltage cascade technology, high-frequency transformers, and medium-voltage insulation technology. A true SST must complete voltage reduction through a single-stage cascade conversion, which imposes extremely high requirements on materials such as silicon carbide and gallium nitride, making insulation difficult. Currently, most SSTs on the market use two-stage or three-stage cascade voltage reduction, increasing the risk of failure. The number of domestic manufacturers capable of producing a single-stage cascade SST is limited, with most still at the prototype development and sample testing stage. The underlying technical routes are largely converging, with differences mainly in scheme design, parameter configuration, and supply chain management. SST system integration is highly complex and challenging to control; grid-connected operation debugging, adaptation, and fault handling are core barriers that rely on long-term engineering experience.

From the perspective of industrialization, the application of solid-state transformers is still in the early demonstration stage, with most companies at a similar progress level and lacking large-scale commercial deployment validation. Sungrow's EnerNeo SST has been officially launched and is available for sale, with plans to establish pilot projects in 2026, begin small-scale deliveries in 2027, and enter mass delivery from 2028 to 2030. The company has signed framework procurement agreements with Zhonglian Data and Dongyangguang. However, telecom operators and major internet companies currently mainly use Panama power supplies for their power distribution, with only about a dozen SST pilot projects nationwide. The scale of signed contracts is not comparable to large computing centers starting at 200 MW capacity.