Nvidia at DAC 2026: Vera CPUs, 50x Faster EDA, and the Strategy of Designing Chips with Its Own Silicon

EDA partners report dramatic performance gains

Through partnerships announced and demonstrated at DAC 2026, Nvidia and its EDA partners reported significant performance improvements across verification, simulation, and thermal management:

• Synopsys unveiled a fully autonomous long-running design verification agent that orchestrates the entire chip verification cycle, delivering up to 50× faster time-to-validated RTL while achieving 20% additional coverage improvement. Synopsys also demonstrated its first fully autonomous computer-aided engineering (CAE) workflow for thermal management and electronic device cooling .
• Cadence demonstrated AI agents built on Nvidia technology that cut chip verification from weeks to hours, using Cadence ChipStack and Nvidia OpenShell .
• Siemens announced an expanded partnership with Nvidia to deliver self-verifying agentic AI workflows for EDA and PCB engineering, moving from autonomous task orchestration toward continuously validated engineering outcomes .

Industry observers noted that DAC 2026 stood out from previous years because performance claims were tied to actual debugging cases, design artifacts, and production tool flows rather than speculative demos .

Agent Toolkit expansion targets chip and systems design

At DAC 2026, Nvidia announced a major expansion of the Nvidia Agent Toolkit specifically targeting electronic and systems design :

• New libraries added: Re-architected Nvidia PhysicsNeMo libraries (giving AI agents physics simulation skills) and updated Nvidia CUDA-X libraries (accelerated solvers and quantum chemistry capabilities) .
• These additions enable developers to build autonomous AI engineers that can plan, optimize, and verify complex chip and system workflows .
• Industry leaders including Cadence, Siemens, and Synopsys are adopting these capabilities into their EDA tools .

Separately, on July 20, Nvidia also added Omniverse libraries to the Agent Toolkit for physical AI capabilities like sensor simulation and GPU-accelerated physics, applicable to chip design and broader engineering .

The broader strategy: eating its own dog food

The overarching theme of Nvidia's DAC 2026 presence was "AI Supercomputing Meets EDA" — the company is now using its own silicon (Vera CPUs + GPUs) and its own AI agent software stack (Agent Toolkit, PhysicsNeMo, CUDA-X) to design its next-generation chips .

Timothy Costa, Nvidia's GM and VP of Computational Engineering, delivered the DAC 2026 Sunday SKYTalk titled "AI Supercomputing Meets EDA: Accelerating the Future of Chip and System Design," framing the strategic vision .

Key strategic elements include:

• Nvidia is using its own products internally: running EDA tools on Vera CPUs and Nvidia GPUs to design the next generation of Vera CPUs and Rubin GPUs .
• The strategy closes a virtuous loop: Nvidia's own AI agents accelerate chip design, and those same chips (Vera, Rubin) are architected to be the best platform for running agentic AI workloads .
• By proving performance gains internally first, Nvidia can credibly sell the same integrated hardware+software stack to the entire semiconductor industry .