AIBRIEF
Two European hardware vendors will pair silicon carbide and gallium nitride power chips with fast discharge supercapacitors, covering both steady state conversion and short power spikes for AI racks.
Infineon and Skeleton Technologies are pairing their respective halves of the power chain that feeds AI data centers, from the wall outlet to the GPU (Compound Semiconductor report).
Under a joint development agreement, the two companies will work on the full path from medium-voltage grid AC to the high-voltage DC an AI rack actually draws. Infineon contributes the power-conversion semiconductors: its CoolSiC silicon-carbide devices for steady-state conversion, and its CoolGaN gallium-nitride switches for the short, intense power spikes that racks pull. Skeleton brings the fast-discharge side: supercapacitors that ride through those spikes in seconds, rather than batteries that hold charge for hours.
The first workstream targets solid-state transformers, meaning solid electronics with no moving parts or iron core, that step grid voltage down to the DC a server rack needs. A second track pairs Infineon's GaN switches with Skeleton's supercapacitors specifically for peak-shaving, smoothing out the brief power surges a dense GPU cluster produces.
Infineon executive Andreas Weisl said the deal responds to data center power density climbing fast enough to force new approaches to conversion, distribution, and storage. Skeleton CEO Taavi Madiberk argued that AI data centers need to deliver more power to denser GPU infrastructure without the power system itself consuming more floor space.
Neither company gave quantitative targets, deployment dates, or named customers, so the agreement is best read as both vendors staking a position on the same engineering constraint, not a product launch.