Integrated circuit and electronic hardware design company Cadence Design Systems Inc. today announced a new artificial intelligence agent that assists with packaging and system design, the step after silicon has been customized and produced, marking how chips and components get used in electronic devices.

AuraStack, a super-agent within the company’s Allegro AI studio, provides agentic design assistance for engineers who are designing and customizing packaging and printed circuit boards that turn silicon into phones, servers and AI infrastructure.

“The automation in system design is pedestrian at best compared to what happens in chip design,” said Vivek Mishra, corporate vice president of the System Design and Analysis Group. “There are a lot of point tools. The teams are quite fragmented. The workflows are tedious.”

Modern electronic products are not simple collections of chips. The components must be combined together with numerous constraints in mind, taking into account power delivery, cooling, wiring and other mechanics on the circuit board that must account for heat, vibration, warping and manufacturing needs. Although chip design has evolved over multiple decades of automation, the packaging for chips (the black plastic square with “pins” coming out of it that people associate with chips on boards) and design of the systems they are placed on still require in-depth analysis, specialized tools and careful design decisions before they enter production.

In a conventional workflow, a board might be laid out before the thermal needs are examined (for example, hotspots); a signal-integrity specialist might then find an electrical problem, or a mechanical simulation might reveal that an important component could loosen under vibration. Every discovery matters and forces extra time and energy into design at later stages.

“Somebody does the PCB, throws it over the wall,” Mishra explained. “Somebody else does the analysis and tells them what to change. Once the design is done, then they look at the mechanical analysis and, ‘Oh, I’ve got to change this, change that.’”

Changing the game for electronics development

AuraStack follows the company’s ChipStack release, an AI agent that assists with silicon design. This agent provides the mental model and super-agent architecture, enabling understanding of the entire system at once, collaboration with engineers, and integration of fragmented procedures into a single work cycle.

According to Mishra, much of the design of PCBs and advanced packaging for chips is still highly assisted rather than fully autonomous. That may come later, but currently it is designed to change the critical work and provide engineers with what they need to fully understand the physics, mechanics, and entire process before settling on a design.

“It makes an average engineer super good and a good engineer into a super engineer,” Mishra said.

Mishra did not argue that AuraStack replaces engineers, rather he said it can broaden the range of analysis available to engineers and reduce the total connective work required to consult multiple teams. Ideally, he said, “What used to take four days can now be done in minutes.”

This is an example of what could be called an agentic full-stack play, assigning a consequential role to AI agents throughout the process of delivering custom silicon all the way to custom boards.

ChipAgents and other startups are pursuing a similar transition toward agents that plan, execute and validate multistep chip-design work. On the printed circuit board side, startups such as Flux are also rolling out AI agents for designing circuit boards, while firms like CircuitHub Inc. are automating PCB manufacturing and assembly at large scale.

Images: SiliconANGLE/Microsoft Designer; Cadence Design Systems