Nvidia released NemoClaw at GTC Taipei at COMPUTEX on June 2, 2026, an open blueprint for building specialized, long-running AI agents aimed at industrial engineering workflows. More than a dozen engineering software providers — including Cadence, Siemens, Synopsys, Dassault Systèmes and Ansys — are using the stack to build autonomous agents that automate computer-aided design, meshing, simulation setup, debugging and post-processing. The pitch is straightforward: accelerated computing already compressed simulation runtimes from weeks to hours, but the surrounding workflow stayed manual. NemoClaw targets that gap.
The blueprint bundles a secure runtime, a model router, frontier models and Nvidia NeMo customization libraries, and integrates with orchestration frameworks including OpenClaw and Hermes. The runtime layer is Nvidia OpenShell, an open source component that governs how each agent touches files, networks and tools, with policy-based security enforced at every layer. Deployment targets span Nvidia DGX Spark personal AI supercomputers, enterprise data centers and cloud providers — covering the workstation-to-cloud range engineering shops actually use.
Cadence is the headline customer. The company is building an autonomous register-transfer level engineer on NemoClaw that orchestrates Cadence Design Systems ChipStack for design and verification. The workflow, demoed in a GTC Taipei keynote, is cutting RTL verification — historically one of the most time-consuming steps in digital circuit design — from weeks to hours. For a chip-design industry where verification often consumes more engineering hours than design itself, a hours-not-weeks loop reshapes how teams schedule tape-outs.
Key facts
- 01Nvidia unveiled NemoClaw at GTC Taipei at COMPUTEX on June 2, 2026, an open blueprint for long-running industrial AI agents.
- 02More than a dozen engineering software providers including Cadence, Siemens, Synopsys and Dassault Systèmes are building agents on the stack.
- 03Cadence's NemoClaw-based RTL engineer with ChipStack is cutting register-transfer level verification from weeks to hours.
- 04NemoClaw deploys on Nvidia DGX Spark personal AI supercomputers as well as enterprise data centers and cloud providers.
- 05PhysicsX is partnering with Microsoft Surface to automate the full thermal simulation lifecycle for Surface laptops.
Siemens is integrating NemoClaw and OpenShell into Fuse EDA AI Agent, an autonomous agent that plans and orchestrates multi-tool workflows across semiconductor, 3D integrated circuit and printed circuit board design. Synopsys is building end-to-end engineering agents on NemoClaw, with Ansys Icepak — part of the Synopsys portfolio — running inside a NemoClaw agent that meshes, simulates and optimizes GPU electronics cooling on the COMPUTEX show floor this week. Dassault Systèmes is productizing the 3DEXPERIENCE Agentic Platform on top of NemoClaw and OpenShell for design, simulation and manufacturing operations.
Startups are filling out the long tail. Flexcompute is applying OpenShell to its Tidy3D and PhotonForge agents for multiphysics co-packaged optics design, combining optical, electrical and thermal simulation to explore thousands of design variants overnight. Nvidia itself is using Flexcompute technology to design and optimize its own optical and photonic devices — a useful tell that the agents are running on production hardware decisions, not just demos.
nTop, the geometry engine behind JetZero's blended-wing-body aircraft program, is using NemoClaw to run autonomous design workflows that compress days of geometry iteration into hours. Neural Concept is deploying an agent for electric motor design that chains electromagnetic, structural and noise-vibration-harness simulations. Luminary is using NemoClaw to autonomously orchestrate data generation, model selection, and training loops for physics AI models. SimScale is targeting hundreds of cross-industry use cases including noise, vibration and harshness analysis — workflows that the company says previously required multiple engineers working over several weeks.
PhysicsX is partnering with the Microsoft Surface team to build an electronics thermal simulation agent that automates the full thermal lifecycle for Surface laptops, combining the PhysicsX platform, Microsoft Discovery and NemoClaw. The agent handles mesh sensitivity analysis, simulation data generation, physics AI model training and optimization-loop execution, with continuous accuracy monitoring across the design exploration. P-1 AI is building Archie, an AI mechanical and electrical engineer already deployed on data center cooling and critical power systems, with automotive, aerospace and national security work next. In a workflow representative of P-1's engagement with Daikin Applied Americas, Archie synthesizes requirements, picks components, runs trade studies and produces engineering artifacts.
The launch fits a broader Nvidia push this week to unify the agentic AI stack across Windows devices, Azure and on-prem environments, which AI Chat Daily covered earlier. NemoClaw is the industrial-engineering vertical of that strategy — a sector where the customers are software incumbents with decades of installed CAD, EDA and CAE tooling, not new SaaS upstarts. By shipping a runtime that plugs into existing orchestration frameworks rather than demanding wholesale replacement, Nvidia is making the integration math easier for those incumbents.
Open questions remain. Nvidia has not disclosed pricing, throughput benchmarks, or any independent evaluation of agent reliability across long-running tasks — and long-horizon reliability is the persistent failure mode for autonomous agents. The vendor demos compress weeks into hours under controlled conditions; production deployments at Cadence-scale customers will reveal how often agents derail mid-workflow and what the human-in-the-loop overhead actually looks like. Security guarantees around OpenShell's policy enforcement also have not been audited publicly.
NemoClaw matters less as a single product launch and more as a positioning move. Nvidia is taking the agentic AI thesis — which most coverage frames around customer-service bots and coding assistants — and pointing it at the highest-margin engineering software market on the planet, where Cadence, Synopsys, Siemens and Dassault Systèmes collectively pull tens of billions in annual revenue. If the agents land, those vendors get to charge for an entirely new tier of automation while running on Nvidia silicon top to bottom. That is the durable revenue compound Nvidia is buying with this blueprint.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
