DAC, the chip industry's main annual engineering gathering, wrapped in Long Beach with Nvidia's first ever booth, AI agents running design steps on their own entering the chip design pipeline, and the new ceiling: simulating heat, electromagnetism,
At DAC 2026 in Long Beach last week, three threads converged: agentic AI, multi-physics simulation, and standards. DAC is the chip industry's main annual engineering gathering, and this year the sub-themes pointed in the same direction.
Nvidia ran its first-ever booth, with VP and GM Tim Costa making the pitch that the limit on product development is no longer how many engineers you can hire, but how much compute you can throw at the problem. EDA, the electronic design automation toolchain that turns a chip spec into a manufacturable design, has spent a year absorbing agentic AI, meaning AI agents that can run steps of the design workflow on their own. At DAC 2025, the talk was about demos. At DAC 2026, a new wave of startups was pitching agent-native tools, and Synopsys floated a token-usage business model on a panel.
Underneath both moves sits multi-physics: thermal, electromagnetic, and mechanical behavior that must be simulated together as chips stack into advanced packages. Wally Rhines, a year into Silvaco, told the show agentic AI will not lift that ceiling. Accellera's functional-safety work and Intel's PADK push point at the lever that might: shared standards, because no single vendor can sit at every table.
What remains open is whether simulation capacity and standards participation move fast enough to keep engineering productivity from stalling. DAC 2027 lands in San Jose.