Ankur Gupta, Siemens | CASPA CEO Summit 2026
summary
Claims that AI halves custom-silicon schedules — 18 months to 9 months with OpenAI and Broadcom — and that EDA must remain deterministic drive the work: Siemens and NVIDIA deploy long-running self-verifying agents to explore 3D IC stacks, thermal and mechanical tradeoffs, and agent-to-agent handoffs across silos via shared data lakes with semantic context. Practical wins include GPU-accelerated EDA, for example Calibre + NVIDIA cuLitho showing a 12x speedup, and end-to-end flows for advanced packaging, co-packaged optics and physical-AI edge emulation.
AI-written summary · about
chapters (8)
- 0:00Training Trustworthy AI for Silicon Design
- 2:35Custom Silicon Demand Requires Faster Design
- 4:57AI-Driven EDA Verification and Design Space Exploration
- 9:03Self-Verifying Agents for 3D Chip Design
- 12:04Connecting Engineering Silos With Shared Context
- 16:03Automating and Accelerating EDA Workflows
- 19:58Design and Emulation for Advanced Computing
- 22:06Holistic Design From Chips to Systems
