onsemi EPP: Wafer-as-Package Power Integration — Up to 3–5× Density Claim, Not Another Discrete SiC Module

 

Bottom line: onsemi introduced the Embedded Power Platform (EPP) on 16 September 2026 as a system-level power architecture that treats the silicon wafer itself as the package foundation. Heterogeneous power dies — silicon, SiC, and GaN — plus FETs, drivers, and controllers can be embedded into cavities on a 12-inch silicon wafer and connected with wafer-level redistribution layers (RDL) rather than wire bonds. Electrical, mechanical, and thermal behavior are co-developed from the start. Official messaging cites up to about 3–5× higher power density versus current solutions, depending on the application. An early EPP-based solid-state circuit-breaker design is described as roughly 50% smaller and 20% cooler. Read this as a wafer-level packaging and power-integration play, not a new discrete SiC MOSFET rating — different from Toshiba IX800- or ROHM-style module news. Subaru is an early engagement partner receiving engineering samples and simulation models; that is evaluation access, not a production award.

What happened

• Architecture: Wafer-as-package on 12-inch silicon; cavity embed of heterogeneous dies; wafer-level RDL replaces wire bonds to cut parasitic inductance.
• Stack: Si, SiC, and GaN can be combined; drivers and controllers ride in the same package.
• Targets: EV 400V/800V architectures; AI data-center 800V racks and dense low-voltage GPU power nodes.
• Partner: Subaru gets early samples and models for future EV architecture evaluation — engagement, not a locked design-win.
• Sources: onsemi press (2026-09-16); Data Center Dynamics; ServeTheHome; Fierce Sensors; Power Systems Design.

Why this is not “another SiC module”: EPP’s thesis turns packaging into an active contributor — full-footprint thermal conduction, package-integrated high-voltage isolation, and multi-physics co-optimization. The 3–5× density figure and SSCB ~50% / ~20% cooler numbers are company claims from early example designs, not independent benchmarks. No part numbers, RDS(on), pricing, or ship dates are invented here.

Four engineering takeaways

1. Wafer-as-package rewrites parasitics first
Pulling interconnect into fab-precision RDL aims for lower inductance and tighter control loops — relevant for 800V traction switching envelopes and compact, telemetrable solid-state breakers at rack scale, even before the SiC-versus-GaN debate.

2. Heterogeneous embed makes materials a configuration choice
Mix Si / SiC / GaN, parallel devices, half-bridge EV layouts, or power-plus-controller in one package. The lever is system-level integration, not a drop-in higher-voltage discrete MOSFET rating.

3. Density claims need a baseline
“Up to 3–5×” is application-dependent. The SSCB example is a directional early datapoint for high-power 800V rack breakers, but without published reference conditions it stays soft-worded as a company claim.

4. Automotive and AI 800V share density pressure, not the same clock
Subaru’s engagement follows long automotive cycles; AI commentary leans into denser 800V DC racks and smaller digital breakers. Same architecture story, different qualification bars — sample access is not “shipping in production.”

Open questions

• What package, cooling, and switching baseline sits under the 3–5× claim?
• How will yield, rework, and high-voltage reliability be proven for mixed-material embeds?
• Where does EPP cross conventional and sintered power modules on cost?
• Will public reference designs appear after the Subaru evaluation phase?

Watch on The Engineering Core

• From 400V to 800V: The Silicon Carbide Revolution — why voltage architecture and SiC moved center stage; EPP is the next layer: packaging and system integration on top of the device story.

Question for you: For the next 800V wave — EV traction or AI racks — do you expect the bottleneck to hit device materials (SiC/GaN) first, or packaging and system-level integration?

Sources: onsemi Embedded Power Platform press announcement (2026-09-16); Data Center Dynamics; ServeTheHome; Fierce Sensors; Power Systems Design. Density and SSCB figures are company claims from early example designs, not independent measurements. Subaru is early technical engagement, not a production win.
Long-form: The Automation Core — https://theautomationcore.blogspot.com · Video: The Engineering Core YouTube channel