AI is arriving in EDA at a time when engineers can no longer depend on brute-force scaling for performance and power improvements, and when the gaps created by disconnected point tools are becoming increasingly time-consuming to fill.

As designs expand to tens of billions of transistors and incorporate chiplets in multi-die assemblies, the hardest problems increasingly sit between tools — preserving intent, managing interdependencies, and identifying potential failures before they reach tapeout. At advanced nodes, where design costs can reach hundreds of millions of dollars and a single respin can threaten an entire program, AI’s real value will depend on whether it can accelerate various steps in the design flow while organizing and interpreting the data needed to trust the results.

The balance is between reliability and practicality. The global shortage of skilled engineers has not gone away. But it has made workflow automation more attractive, provided it can either improve the productivity of human engineers or achieve comparable results.

“Every customer is saying, ‘Give me the ability to shrink my time to results by at least 2×,’” said Ankur Gupta, executive vice president and head of EDA IC software at Siemens EDA. “Demand is pushing EDA toward a new operating model built on faster engines, smarter execution, and trusted outcomes. Acceleration may come from GPUs, machine learning, or reinforcement learning, but the larger goal is to shorten design cycles without losing confidence in the result.”

Efficiency also matters at the agent layer. Allowing any agent to call any tool may be flexible, but it can be expensive and inefficient if the underlying interfaces are poorly written.

Source link