Onsemi Aims to Shrink Power Systems for EVs and AI

A new semiconductor platform promises to pack significantly more electrical power into smaller spaces, potentially altering the physical design of electric vehicles and data centers.
Onsemi has introduced a new power integration architecture designed to address the growing spatial and thermal constraints in electric vehicles and artificial intelligence infrastructure. The company claims this embedded platform can deliver three to five times higher power density than current standard solutions, allowing engineers to fit more electrical capacity into a smaller physical footprint.
According to GN auto tech/ev, this approach is particularly relevant as both EVs and AI data centers face strict limits on space and heat dissipation. By combining multiple power devices within a single silicon-based package, the platform seeks to reduce design complexity and improve overall system efficiency without requiring larger chassis or cooling systems.
Integrating diverse semiconductor materials
The core of the platform is its ability to combine silicon, silicon carbide, and gallium nitride devices alongside drivers and controllers in configurable combinations. These materials occupy different roles in power electronics, with silicon carbide and gallium nitride typically handling higher voltage and faster switching tasks. Onsemi positions this architecture as a cohesive system rather than a collection of separate components pieced together later, allowing for tailored solutions based on specific power needs.
This multi-material integration is a significant shift from traditional approaches where each device type might be treated independently. By optimizing electrical, thermal, and mechanical performance from the start of the design process, the company aims to prevent the common engineering headache of reconciling conflicting requirements late in development. Subaru has partnered with Onsemi to explore early applications in electrified vehicles and dense AI data centers, signaling industry interest in this integrated approach.
Replacing traditional wire bonds
A key technical distinction in this platform is the use of wafer-level redistribution layers instead of traditional wire bonds for internal electrical connections. Wire bonds have long been the standard method for connecting chips inside packages, but Onsemi argues that redistribution layers offer a more direct and controlled pathway. This design change is intended to reduce parasitic inductance, which is an unwanted delay in energy transfer, thereby supporting faster switching frequencies and potentially more reliable device performance.
The shift away from wire bonds also impacts the development workflow. Onsemi states that the architecture supports shorter development cycles through simulation and digital-twin modeling. This allows engineers to test and optimize designs earlier in the process, reducing the need for physical prototyping iterations. However, the practical benefits of these density gains depend heavily on the specific system implementation and how effectively the thermal management is handled in real-world conditions.
Trade-offs and practical implications
While higher power density offers clear advantages for miniaturization, it is not a universal solution. The practical results of using such high-density platforms are contingent on the specific system where the technology is deployed. Engineers must still manage the intense heat generated by packing more power into a smaller area. The platform’s value lies in its integrated approach to managing these trade-offs, but it requires a holistic design strategy to fully realize the promised efficiency gains.
For manufacturers, this means a shift in how power architecture is viewed. It is no longer just a component-level decision but a factor that influences the overall design of a vehicle or data-center rack. As the demand for efficient power delivery continues to rise, platforms like Onsemi’s embedded architecture could become standard in high-performance electronics, though widespread adoption will depend on cost and compatibility with existing manufacturing processes.






