globenewswire.com web signal

NVIDIA Launches Enterprise Agent Toolkit With SAP, Palantir

Key insights

  • Nemotron 3 Ultra, a 550-billion-parameter mixture-of-experts model, delivers 5x faster inference and up to 30% lower cost versus comparable open models.
  • Cadence, Siemens, Synopsys, and Dassault Systèmes are deploying autonomous AI engineers using NVIDIA's NemoClaw blueprints.
  • Microsoft, Canonical, Red Hat, SAP, and ServiceNow are integrating OpenShell for policy-based enterprise agent management across desktops, servers, and cloud.

Why this matters

Eleven named enterprise software leaders spanning EDA, cybersecurity, operations, and cloud infrastructure are adopting the toolkit simultaneously, signaling that enterprise AI agents are moving from experimentation into production deployment. OpenShell's policy-based runtime directly addresses the governance gap that has blocked most enterprises from deploying autonomous agents in regulated or multi-vendor environments. The 5x inference speed and 30% cost reduction from Nemotron 3 Ultra shifts the economics of large-parameter agentic models, making continuous enterprise workloads viable rather than prohibitively expensive.

Summary

NVIDIA introduced its Agent Toolkit at GTC Taipei on June 1, bundling NemoClaw blueprints, Nemotron open models, OpenShell secure runtime, and CUDA-X libraries into a platform for enterprise autonomous AI agent deployments. Nemotron 3 Ultra, a 550-billion-parameter mixture-of-experts model, delivers 5x faster inference and up to 30% lower cost versus comparable open models, with native support for LangChain Deep Agents and OpenHands. Essentially: (Cadence, Siemens, Synopsys, Dassault Systèmes) deploy autonomous AI engineers via NemoClaw; CrowdStrike uses Nemotron for vulnerability identification and remediation; Palantir embeds it for autonomous task execution. - Microsoft, Canonical, Red Hat, SAP, and ServiceNow integrate OpenShell for policy-based agent management across desktops, servers, and cloud environments. - Nemotron 3 Ultra supports LangChain Deep Agents and OpenHands as named agent platforms. The announcement marks NVIDIA's push to own the software infrastructure layer of enterprise agentic AI, well beyond its hardware role.

Potential risks and opportunities

Risks

  • CrowdStrike's reliance on Nemotron for vulnerability identification and remediation creates a single-vendor dependency; a model regression or adversarial prompt injection could introduce security gaps in production environments.
  • Enterprises integrating OpenShell across Microsoft, Canonical, Red Hat, SAP, and ServiceNow face policy interoperability risk if each vendor implements controls differently, fragmenting governance.
  • Palantir's autonomous task execution via Nemotron raises unresolved liability questions if agents act erroneously in enterprise workflows, with no attribution framework between NVIDIA and Palantir publicly disclosed.

Opportunities

  • LangChain and OpenHands, both named as supported platforms for Nemotron 3 Ultra, gain immediate enterprise distribution through NVIDIA's partner network without additional integration overhead.
  • Microsoft, Red Hat, Canonical, SAP, and ServiceNow each gain a differentiated AI governance offering by embedding OpenShell, potentially accelerating enterprise AI contract wins against cloud-native competitors.
  • Enterprises already licensed on Cadence, Siemens, Synopsys, or Dassault Systèmes EDA tools can begin piloting autonomous AI engineers immediately using a vetted, partner-supported toolkit.

What we don't know yet

  • No pricing or licensing terms disclosed for Agent Toolkit components or enterprise access to Nemotron 3 Ultra as of June 1.
  • Whether the autonomous AI engineers deployments at Cadence, Siemens, Synopsys, and Dassault Systèmes are in production or still in early access pilot.
  • How OpenShell enforces consistent policy across Microsoft, Red Hat, SAP, and ServiceNow environments, and whether it requires NVIDIA hardware to operate.