Santa Clara, Calif. | EcoPulse24

NVIDIA announced today a major expansion of its NVIDIA Agent Toolkit for engineering, adding re-architected NVIDIA PhysicsNeMo™ libraries and updated NVIDIA CUDA-X™ libraries as agent-ready tools designed to transform how products, chips and complex engineering systems are designed and developed.

The company said the expanded toolkit enables software developers to build autonomous AI engineers capable of applying AI physics skills, running accelerated engineering solvers and performing quantum chemistry simulations across increasingly complex engineering workflows.

NVIDIA also announced that NVIDIA Nemotron™ 3 Ultra now leads among open models in agentic register-transfer level (RTL) coding using the ACE-RTL hardware-design agent developed by NVIDIA Research, providing enterprises with a customizable foundation for AI-powered chip design and verification.

According to NVIDIA, industry leaders including Cadence, Siemens, Synopsys and other engineering software companies are already deploying NVIDIA accelerated computing and agentic AI technologies to advance autonomous engineering workflows spanning chip architecture, verification, advanced packaging and complete system development.

AI Engineers Designed for Modern Engineering Complexity

NVIDIA said designing the next generation of semiconductor devices and computing systems increasingly requires engineering teams to connect physics, simulation and performance analysis across highly complex design cycles.

To address that challenge, the company said a new generation of autonomous AI engineers is emerging that can operate specialized engineering tools, execute simulations and generate high-fidelity engineering data to help scale chip design, verification, packaging and systems engineering.

To support that transition, NVIDIA has re-architected PhysicsNeMo into agent-friendly software libraries and expanded the Agent Toolkit with new and updated CUDA-X libraries.

PhysicsNeMo provides AI physics capabilities for training and deploying physics-based AI models, while CUDA-X introduces accelerated engineering solvers together with quantum chemistry capabilities into agentic engineering workflows.

"Engineering has reached an inflection point. AI can now work with tools of physics, simulation and design," said Timothy Costa, Vice President and General Manager of Computational Engineering at NVIDIA.

"With NVIDIA Agent Toolkit, developers can build agentic engineers that reason using physics, run complex simulations and generate high-fidelity data to become a new engine for innovation in chip and system design."

NVIDIA Agent Toolkit Adds AI Physics and Accelerated Computing

NVIDIA said the Agent Toolkit enables developers to build specialized engineering AI assistants connected to domain-specific models, engineering tools and enterprise data.

With the addition of PhysicsNeMo and CUDA-X libraries, engineering agents gain access to AI physics models, accelerated numerical solvers and quantum chemistry capabilities for semiconductor, systems and industrial engineering.

What is NVIDIA Agent Toolkit?

NVIDIA Agent Toolkit is a software development framework that enables developers to build specialized AI agents capable of performing engineering tasks rather than simply answering questions. These AI agents can connect to engineering software, simulation tools, AI models and enterprise data to assist with chip design, verification, industrial engineering and scientific computing workflows.

With the latest expansion, the toolkit now incorporates NVIDIA PhysicsNeMo™ libraries for AI physics modeling and updated NVIDIA CUDA-X™ libraries, allowing AI agents to apply physics-based reasoning, execute accelerated numerical simulations and perform quantum chemistry calculations. According to NVIDIA, the platform is designed to help developers build autonomous AI engineers that can support increasingly complex engineering and semiconductor design processes.

AI Physics Skills

PhysicsNeMo libraries allow AI agents to train and deploy customizable AI physics models for complex simulation and engineering tasks while converting model architectures into callable tools that can be integrated into engineering workflows.

Iterative Sparse Solvers

The newly introduced NVIDIA cuISS (CUDA Iterative Sparse Solvers) library accelerates large sparse linear systems commonly used in engineering and physics simulations.

NVIDIA said the library is optimized for GPU flexibility and performance through modern composable solvers and preconditioners that enable developers to build scalable production-grade simulation engines for autonomous engineering workflows.

Direct Sparse Solvers

The NVIDIA cuDSS (CUDA Direct Sparse Solvers) library accelerates complex sparse linear systems that are central to electronic design automation (EDA) and scientific simulation.

The company said cuDSS provides high numerical robustness and performance for device, circuit and system simulations while scaling across multi-GPU and multi-node production deployments.

Quantum Chemistry

The new NVIDIA cuEST (CUDA Electronic Structure Theory) library brings high-accuracy quantum chemistry simulations to device-relevant scales.

According to NVIDIA, cuEST enables both density functional theory (DFT) and post-DFT methods to be integrated into production engineering workflows while supporting a broad range of modern functionals and making increasingly large ground-state and excited-state simulations practical on NVIDIA GPUs.

Nemotron 3 Ultra Advances AI Coding for Semiconductor Design

NVIDIA said semiconductor development depends on specialized register-transfer level (RTL) coding that requires high accuracy, deep engineering expertise and deployment flexibility.

Using ACE-RTL, an autonomous hardware-design agent developed by NVIDIA Research, NVIDIA Nemotron 3 Ultra now leads among open models in agentic RTL coding across the comprehensive Verilog design problems benchmark.

The company said Nemotron 3 Ultra combines industry-leading accuracy and efficiency while allowing enterprises to post-train models using proprietary datasets and deploy them locally or on premises to maintain greater customization, operational control and data privacy.

Developers can begin using Nemotron 3 Ultra through:

• Cadence's harness

• Synopsys' fully autonomous long-running design verification and analog/mixed-signal agents

• Siemens' Questa One smart verification agentic toolkit

• Hugging Face

Engineering Software Leaders Expand Autonomous AI Workflows

NVIDIA said major industrial engineering software companies are already integrating the expanded Agent Toolkit into production engineering platforms.

Cadence

Cadence is integrating NVIDIA Nemotron, accelerated computing and CUDA-X libraries with the recently introduced Cadence AuraStack AI Super Agent and the Cadence Millennium M2000 platform.

According to NVIDIA, the platform autonomously manages advanced packaging and printed circuit board (PCB) design from design exploration through manufacturing signoff while delivering up to 20x faster multiphysics performance.

Cadence's silicon design super-agent portfolio now spans the semiconductor workflow from architecture through manufacturing signoff.

The collaboration also extends to infrastructure optimization, with Cadence's EDA and system design automation portfolio - including the Cadence Jasper formal verification platform - being optimized for the NVIDIA Vera CPU to accelerate advanced chip validation.

Synopsys

Synopsys is using the NVIDIA Agent Toolkit together with NVIDIA NIM™ microservices, Nemotron open models, the NVIDIA NeMo™ Gym library and NVIDIA NemoClaw™ blueprints inside Synopsys AgentEngineer to create secure accelerated AI workflows across semiconductor and system design.

Using Ansys Icepak, Synopsys' autonomous engineering workflow performs simulation setup together with pre-processing and post-processing for GPU cooling optimization.

The company is also developing engineering use cases around NVIDIA cuISS to accelerate simulation workloads.

Beyond software workflows, Synopsys VCS, the company's functional verification platform, is being optimized for the NVIDIA Vera CPU to improve verification throughput before chip fabrication.

Siemens

Siemens is combining NVIDIA NeMo Gym, Nemotron open models and CUDA-X libraries with the Siemens Fuse EDA AI Agent to orchestrate multi-tool and multi-agent engineering workflows across semiconductor, 3D-IC, PCB and complete system design - from concept through manufacturing signoff.

Within the Siemens Solido Characterization Suite, NVIDIA said these AI workflows are delivering more than 10x faster library characterization while reducing token costs by more than 10x.

Samsung

Samsung is deploying NVIDIA cuLitho together with CUDA-X libraries to achieve up to 20x greater computational lithography performance.

The company is also applying NVIDIA PhysicsNeMo to perform chip-scale thermal-stress analysis with numerical solver-level accuracy across engineering domains containing up to 10 billion cells.

ChipAgents

ChipAgents is using the NVIDIA Agent Toolkit to develop domain-specific AI agents for semiconductor design and verification.

The company is fine-tuning NVIDIA Nemotron models for complete semiconductor engineering workflows including debugging, formal verification, coverage analysis and additional verification tasks.

Silvaco

Silvaco is leveraging NVIDIA accelerated computing to scale high-accuracy 3D optical simulation within Silvaco Victory Device.

Running on 32 NVIDIA GPUs connected through NVIDIA NVLink™, the platform completed a 3.2-billion-mesh-node photonic edge coupler simulation in under four hours, a workload NVIDIA said extends beyond the practical limits of CPU-based simulation.

Keysight, Samsung, Synopsys and TSMC

Keysight is using NVIDIA cuDSS to accelerate electromagnetic simulations by up to 10x.

Meanwhile, Samsung, Synopsys and TSMC are integrating NVIDIA cuEST into GPU-accelerated engineering pipelines to achieve up to a 50x speedup for key quantum chemistry workloads.

NVIDIA invited developers and industry professionals to learn more about the expanded engineering platform during the Design Automation Conference (DAC).

Key Announcements

EcoPulse24 Analysis

NVIDIA's latest expansion of the Agent Toolkit reflects a strategic shift beyond generative AI assistants toward autonomous engineering agents capable of executing specialized design, simulation and verification tasks across semiconductor and industrial engineering workflows.

By integrating AI physics, accelerated numerical solvers and quantum chemistry capabilities into a unified agent framework, NVIDIA is positioning its accelerated computing platform at the center of next-generation electronic design automation (EDA). The broad adoption by leading software vendors - including Cadence, Siemens, Synopsys and Samsung - signals growing industry momentum toward AI-driven engineering pipelines that combine domain expertise, high-performance computing and autonomous decision-making to reduce development time for increasingly complex chips and systems.