onsemi Embedded Power Platform — AI Data Center Power Portfolio

Specifications

Overview

The onsemi Embedded Power Platform (EPP) is a power packaging architecture announced on September 16, 2026 that turns the silicon wafer itself into the package rather than treating package and die as separate technologies. Multiple devices — FETs, gate drivers and controllers — are embedded together and co-optimised for electrical, thermal and mechanical behaviour, which onsemi rates at up to 3-5x the power density of conventional power solutions.

EPP mixes silicon, silicon carbide and gallium nitride inside the same highly integrated wafer-level architecture and relies on standard 12-inch silicon wafer manufacturing, bringing integration processes under fab-level process control. The platform begins sampling with strategic customers in 2026 across automotive and AI applications; Subaru is an early engagement partner evaluating EPP for future electrified vehicle architectures.

For AI data centers specifically, EPP sits on top of onsemi's existing high-voltage DC work. onsemi is collaborating with NVIDIA on the transition to 800 VDC rack power architecture, supplying solid state transformers, power supply units, 800 VDC distribution and core power delivery with integrated monitoring and control. The 800 VDC approach allows direct conversion from 13.8 kV AC to 800 V, cutting conversion stages and losses compared with the 12 V and 48 V rails used in current racks.

Key Benefits

Density where it matters — up to 3-5x higher power density addresses the space constraint in AI racks where compute, not power, is expected to fill the chassis. One co-designed platform — electrical, thermal and mechanical behaviour are modelled together instead of sequentially, reducing late-stage rework. Fab-grade integration — 12-inch wafer-level manufacturing applies semiconductor process control to power system assembly. Aligned to 800 VDC — EPP is aimed at the high-voltage DC architecture NVIDIA has signalled for next-generation AI data centers.

Applications

AI data center power delivery and 800 VDC rack distribution, solid state transformers and grid-to-GPU power chains, core power delivery for high-power GPU platforms, automotive electrification including traction and DC-DC conversion, and industrial automation systems competing for the same power budget per litre.

Related Products