Socionext is adopting Cadence's AI agent for SoC design using a 3nm process. It aims to automate physical design and increase engineer productivity by up to 5x.
On 20 October 2026, Socionext announced that it would deploy the US-based Cadence's AI agent technology, InnoStack AI Super Agent, for the physical design of large-scale SoCs using an advanced 3nm process. By strengthening the automation of design workflows, the company aims to increase engineer productivity by up to 5x.
As mass production of 3nm-generation products advances, the complexity of physical design has increased substantially. In particular, the development of custom SoCs for AI data centers and HPC requires enormous amounts of engineering effort and advanced specialist knowledge. Against this backdrop, EDA vendors are expanding their offerings of agentic AI, which strengthens generative AI by referencing external knowledge.
Socionext will deploy Cadence's InnoStack AI Super Agent, using technologies such as RAG and MCP to integrate in-house design knowledge into physical design workflows. The targeted areas include floorplanning, timing-analysis checks and, in the future, checklist automation. President Yoshida Hisato (吉田 久人) said the company expects overall productivity to improve by up to 5x.
Going forward, the company plans to link the use of AI in front-end and physical design and accumulate and reuse design methodologies as organizational assets. This will enable experienced design expertise to be used across the organization through AI agents, with the company aiming to establish a new standard model for next-generation custom SoC development.