ChipAgents reposted this
I had the privilege of moderating the #GSA Tech Summit panel “Accelerating Design-to-Fabrication – AI’s new frontier?” on May 21, 2026, in Scottsdale, Arizona, I was joined by an exceptional group of leaders: Ahmed Shariq, Distinguished Product & Quality Engineer at Ampere; Arun Venkatachar, Vice President, Artificial Intelligence, Cloud & Central Engineering at Synopsys Inc ; Paul Cunningham, Senior Vice President & General Manager, System Verification Group at Cadence; and William Yang Wang, Founder & CEO of ChipAgents. We went deep on what AI-native EDA actually looks like in production: semi-autonomous agents evolving from simple copilots to multi-agent systems that compress weeks of work into hours, while still respecting the hard boundaries of deterministic, physics-based signoff. A key idea was the “layer cake” of agentic AI in chip design—design teams building their own agents for local automation, workflow orchestration layers spanning heterogeneous tools, and deep engine-embedded agents inside EDA stacks that truly optimize PPA and convergence, not just generate suggestions. We also discussed how expert-system structure plus #LLM can form a kind of “virtual agent organism,” particularly powerful in verification where probabilistic methods can augment, but never replace, deterministic #STA and #signoff. Infrastructure and cost emerged as critical design parameters, not afterthoughts. Running AI for chip design at scale means thinking hard about CPU/GPU balance, how you orchestrate heterogeneous compute, and how you keep a lid on 20x #token growth as LLM-based flows expand. We talked about infra-level #cost efficiency, open-source tool ecosystems, and why “Internet-open” environments plus strong #governance are essential to avoid building a Ferrari stack you only use to get groceries. Finally, we tackled the design-to-fab bridge: how process-aware design to maximize yield, #secure use of #fab data, and patterns like federated learning (borrowed from industries like banking) can help break historic silos between #fabless companies, foundries, and EDA vendors. My takeaway: AI may be the only practical way to close this loop at scale—enabling pragmatic, multiphysics-aware optimization, opening the door to zero‑revision NPIs, and even “virtual chip companies” operating on top of rich, AI-orchestrated compute and tooling stacks by 2029. Want to learn how AWS is empowering Semiconductor companies build and deploy agentic AI flows for chip design and infrastructure at scale? DM me. Thank you to GSA and #Deloitte for making this event successful Gautham Unni, PhD Sarah Cooper Dhara Vaishnav Kathleen Mason Louis Kalmar Beko Jang Amanda Lundy Alex Kupres Avery Lu Kole Giles #aws #GSA #WhereLeadersMeet #semiconductor #EDA #AI #chipdesign #cloud #infrastructure #agents #manufacturing
-
-
-
-
-
+2