Summary
- Infineon and Skeleton have signed an MoU covering solid-state transformers and high-power sidecars for AI data centres.
- The proposed systems combine Infineon SiC and GaN semiconductors with Skeleton power-conversion systems and supercapacitors.
- The work targets higher power density, grid-to-core efficiency, peak management, and resilience rather than conventional lower-density distribution architectures.
Infineon Technologies and Skeleton Technologies have agreed to collaborate on solid-state transformers and high-power sidecar systems for AI data centres, targeting the electrical conversion and peak-power problems created by increasingly dense compute infrastructure.
The companies have signed a memorandum of understanding covering power architecture across the chain from the grid to the processor. Their planned work includes next-generation solid-state transformers for medium-voltage AC to high-voltage DC conversion and peak-shaving systems combining supercapacitors with gallium nitride power semiconductors.
Infineon will contribute its high-voltage CoolSiC silicon carbide and CoolGaN gallium nitride semiconductor technologies. Skeleton will contribute its supercapacitors, power-conversion systems, and high-power energy-storage expertise.
The agreement is a development collaboration rather than a disclosed commercial deployment. No prototype timetable, customer project, capacity, or product launch date has been announced.
Its focus nevertheless sits directly inside one of the engineering constraints created by AI infrastructure: moving much larger quantities of electricity through a finite building footprint without allowing conversion equipment, energy storage, and electrical losses to expand at the same rate as compute demand.
Reworking the grid-to-core chain
Conventional data centre electrical systems pass power through several conversion and distribution stages before electricity reaches server hardware. At high facility loads, losses at each stage become material because even small percentage differences are multiplied across tens or hundreds of megawatts of continuous demand.
Infineon and Skeleton intend to explore solid-state transformers for converting medium-voltage alternating current into high-voltage direct current. The companies say the architecture could improve efficiency and power density while supporting greater resilience across the power chain.
Solid-state transformers use semiconductor-based conversion rather than relying solely on conventional transformer architecture. Their potential advantages include tighter control over power flow and the opportunity to combine voltage conversion with DC distribution and storage, although practical data centre deployment also has to satisfy protection, maintainability, thermal, reliability, and cost requirements.
The companies have not disclosed efficiency targets or equipment capacities, so the scale of any improvement cannot yet be quantified.
The second part of the collaboration focuses on fast changes in electrical demand. Large GPU systems can present rapidly varying loads as compute clusters move between operating states. Skeleton and Infineon plan to develop GaN-based peak-shaving systems combining supercapacitors with power electronics to absorb or deliver high power over short periods.
Supercapacitors serve a different role from long-duration batteries. They are designed around high charge and discharge rates and very large numbers of cycles rather than storing energy for lengthy outages. In a data centre power system, that can make them useful for smoothing short-duration load swings rather than replacing standby generation or batteries intended to sustain operations for longer periods.
Power density drives component change
The collaboration comes as power-system design is becoming more closely coupled to server density. Higher rack loads place pressure on transformers, switchgear, busways, UPS systems, power-conversion stages, and cooling infrastructure, while operators are also trying to limit the amount of plant space consumed by electrical equipment.
Infineon’s semiconductor portfolio already spans several parts of that chain. Skeleton adds high-power storage and system-level power conversion, creating a collaboration focused on the interface between medium-voltage distribution, DC architectures, and rapidly changing compute demand.
DataCentral reported on 2 September that Turbo Power Systems is also expanding its data centre power-conversion work, including bidirectional systems and DC microgrid capability. Both developments illustrate how the AI infrastructure market is drawing electrical suppliers into areas beyond traditional UPS equipment.
The engineering threshold for adoption is high. Medium-voltage distribution infrastructure is expected to remain operational for years and must be serviceable, protectable, and predictable under fault conditions. Operators are unlikely to adopt a denser architecture solely because individual semiconductor devices can switch more efficiently.
Infineon and Skeleton will therefore have to demonstrate system-level gains in efficiency, space, reliability, and operating cost rather than component performance alone.
The MoU does not establish when those systems will reach customer sites, but it places solid-state conversion and high-power short-duration storage among the technologies being evaluated as AI data centres push more electricity through the path between grid connection and processors.
