Get the weekly research letter
Companies change their terms quietly. We read every version and catch what actually changed. One email a week on the changes that matter and what they mean. No account.
The agreement restricts NVIDIA SDKs (including CUDA toolkit, TensorRT, and cuDNN), NVIDIA drivers, and NVIDIA Omniverse to use solely on systems with NVIDIA Platforms, prohibiting their use on non-NVIDIA hardware.
This analysis describes what NVIDIA NIM's agreement states, permits, or reserves. It does not constitute a legal determination about enforceability. Regulatory applicability and practical outcomes may vary by jurisdiction, enforcement context, and individual circumstances. Read our methodology
This provision establishes a binding hardware restriction on core NVIDIA developer and simulation tools, including the widely used CUDA toolkit. Customers who operate heterogeneous computing environments combining NVIDIA and non-NVIDIA accelerators should evaluate whether any workflows involving these SDKs or Omniverse cross into non-NVIDIA hardware contexts.
Under this clause, NVIDIA SDKs such as CUDA toolkit, TensorRT, and cuDNN, as well as NVIDIA drivers and Omniverse, are licensed for use only on systems running NVIDIA Platforms; use of these components on non-NVIDIA hardware would fall outside the permitted license scope.
Cross-platform context
See how other platforms handle NVIDIA Platform Requirement for SDKs, Drivers, and Omniverse and similar clauses.
Compare across platforms →Monitoring
NVIDIA NIM has changed this document before.
Receive same-day alerts, structured change summaries, and monitoring for up to 20 platforms.
"8.15 NVIDIA Platform Requirement. The following proprietary Software components are licensed solely to run on systems with NVIDIA Platforms: (a) NVIDIA software development kits (such as CUDA toolkit, TensorRT, cuDNN), (b) NVIDIA drivers, and (c) NVIDIA Omniverse.Excerpt from NVIDIA NIM's NVIDIA AI Foundation Models AUP
(1) REGULATORY LANDSCAPE: This provision is a commercial hardware restriction and does not directly engage privacy or data protection frameworks. Competition law authorities in the EU (European Commission, DG COMP) and potentially the U.S. Department of Justice may have interest in hardware tying arrangements in AI software markets, though whether this specific clause raises competition law concerns depends on market context and applicable precedent that this document alone does not establish. (2) GOVERNANCE EXPOSURE: Medium. Organizations running multi-vendor GPU environments (for example, combining NVIDIA and AMD or Intel accelerators) must ensure that CUDA toolkit, TensorRT, cuDNN, and Omniverse workflows are confined to NVIDIA hardware. Violations could affect license standing under the master agreement. (3) JURISDICTION FLAGS: The restriction applies globally as written. EU competition law may be relevant context for legal teams evaluating the broader implications of hardware tying in AI software licensing, though this document does not establish a legal conclusion on that question. (4) CONTRACT AND VENDOR IMPLICATIONS: Enterprise customers using cloud or colocation providers should confirm that workloads involving CUDA toolkit, TensorRT, cuDNN, or Omniverse are routed exclusively to NVIDIA GPU instances. Vendor agreements with cloud providers should be reviewed to ensure NVIDIA Platform compliance is maintained. (5) COMPLIANCE CONSIDERATIONS: IT and DevOps teams should audit infrastructure configurations to confirm NVIDIA SDK and Omniverse workloads are not inadvertently scheduled on non-NVIDIA accelerators in heterogeneous cluster environments. This is particularly relevant for Kubernetes or container orchestration environments where workload scheduling may not automatically enforce hardware constraints.
This provision establishes a binding hardware restriction on core NVIDIA developer and simulation tools, including the widely used CUDA toolkit. Customers who operate heterogeneous computing environments combining NVIDIA and non-NVIDIA accelerators should evaluate whether any workflows involving these SDKs or Omniverse cross into non-NVIDIA hardware contexts.
Under this clause, NVIDIA SDKs such as CUDA toolkit, TensorRT, and cuDNN, as well as NVIDIA drivers and Omniverse, are licensed for use only on systems running NVIDIA Platforms; use of these components on non-NVIDIA hardware would fall outside the permitted license scope.
No. ConductAtlas is an independent monitoring service. We are not affiliated with, endorsed by, or sponsored by NVIDIA NIM.