Socionext announced that it has applied Cadence’s AI agent to the physical design of large-scale SoCs using a 3nm process and plans to extend its use to other projects. Overall, it expects productivity to increase by approximately up to 5-fold.
Socionext has worked with Cadence to expand the application of AI agents from front-end design to back-end physical design. The target is the physical design of large-scale SoCs using an advanced 3nm process, with Cadence’s InnoStack AI Super Agent being used. The company plans to extend the technology to other SoC development projects.
InnoStack AI Super Agent is an AI agent for digital implementation and signoff that Cadence announced in April 2026. It covers design processes including logic synthesis, placement and routing, timing analysis and ECOs. Socionext will use RAG, MCP and customized AI skills to incorporate its accumulated design knowledge into physical-design workflows.
Applications include floorplanning, proprietary timing-analysis checks and, in the future, automation of design checklists. The company said it intends to make the design knowledge of experienced engineers available through AI agents and promote reuse across the organization. President Hisato Yoshida (吉田 久人) expects productivity to increase by up to 5-fold overall, although specific evaluation results have not been disclosed.
Cadence has successively rolled out its ChipStack, ViraStack and InnoStack AI agents, while collaboration with Japanese organizations including Honda Research Institute Japan and Rapidus is also expanding. Socionext’s initiative represents an example of Agentic AI beginning to enter actual design workflows for advanced SoCs.