Cadence shipped AuraStack, an AI super agent that coordinates domain-specific agents across PCB and advanced packaging design. RTL, synthesis, and physical signoff have had AI copilots for two years. PCB and packaging (where electrical, mechanical, and thermal constraints actually intersect) has had none until now. The claims are 15X productivity and 2X time-to-market, which if they hold at scale represent a step change in a workflow that has been stubbornly sequential.
The constraint AuraStack removes is not routing speed. It is multiphysics co-analysis being a sequential, expert-gated handoff: layout engineers pass to SI/PI engineers pass to thermal engineers, each running separate tools, each producing results that may require design rework from the previous step. AuraStack puts agent orchestration at that coordination seam, planning, implementation, and multiphysics analysis running in coordinated parallel rather than strict sequence. The Rapidus partnership announced the same week (integrating Cadence's InnoStack AI Super Agent into Rapidus' Raads SoC design environment) confirms this is not just a product launch. Cadence is positioning agent orchestration as the integration layer between design intent and process-specific manufacturing requirements.
PCB and advanced packaging have been the last major EDA workflows resisting programmability. Formats are still largely proprietary and binary, tools are GUI-first, and cross-domain analysis requires specialist handoff that neither open-source nor smaller vendors have been able to automate. If AuraStack's numbers hold, the competitive pressure on PCB-only incumbents (Altium, Zuken) and on internal expert workflows becomes acute within 12-18 months. The more interesting question is whether AuraStack's architecture (agent orchestration over domain-specific tools) becomes the template that displaces the tool itself, or whether it remains a Cadence-platform feature that only benefits teams already in the Cadence stack.