NVIDIA Software Licensing Overview and FAQ

NVIDIA provides a comprehensive platform of open software frameworks, libraries, and microservices designed to accelerate AI development and production workloads. This overview outlines the key components of our licensing model to help you navigate our software offerings, support policies, and legal agreements.

1. Licensing Categories: Open Source, Community and Enterprise

NVIDIA offers three primary types of software based on use case and support requirements:

  • Open Source Software (Free): Much of NVIDIA's software is open source. These products are licensed under standard OSI-approved licenses such as Apache 2.0, BSD, or MIT. Examples include NVIDIA Triton, NVIDIA Dynamo, NVIDIA NeMo, NVIDIA OpenShell and NVIDIA RAPIDS, which are available for download on GitHub and other public open-source repositories.

  • Community Products (Free): These include free SDKs and libraries packaged as microservices in containers and made available in the general NGC catalog. Examples include NVIDIA Triton containers, NVIDIA NeMo containers, and NVIDIA Omniverse. They can be used for both development and production and receive community-level support (e.g., via forums).

  • Enterprise Products (Paid): These are paid software subscriptions, such as NVIDIA AI Enterprise and NVIDIA Mission Control. Enterprise Products include enterprise support.

2. The NVIDIA Software License Agreement

The NVIDIA Software License Agreement serves as the main agreement for Community Products and Enterprise Products. It establishes the general terms of use, including:

  • License Grant: Permission to install, use, and redistribute the software according to your order parameters.

  • Ownership and IP: NVIDIA retains all rights to the software; users receive usage permissions only.

  • Key Restrictions: Users may not reverse engineer, decompile, or remove proprietary notices from the software.

NVIDIA CUDA Toolkit, NVIDIA cuDNN and various other NVIDIA SDKs are licensed under separate SDK license agreements, which are officially documented on the NVIDIA Documentation Hub.

3. Product-Specific Terms & Addendums

To address the unique needs of different technologies, NVIDIA uses Product-Specific Terms as addendums to the main Software License Agreement to modify the main terms as necessary.

  • AI Products: Consolidates terms for both free Community Products, and paid software subscriptions for AI, such as NVIDIA AI Enterprise, and NVIDIA Mission Control. These terms now also apply to Omniverse software.

  • vGPU Products: Covers NVIDIA vWS, NVIDIA vPC and NVIDIA vApps.

  • Networking: Covers NVIDIA DOCA, NVIDIA UFM, and NVIDIA NetQ.

  • Confidential Computing: Specific terms for remote attestation and trust services.

4. Key Licensing Concepts

Understanding these terms is essential for managing your NVIDIA software subscription portfolio:

Term Definition
Subscription

A fixed-duration license (e.g., 1 or 3 years) that includes software access and technical support.

License Program

Grants defined by use case: Production (deployment), Evaluation (90-day trial), Developer (development and testing only), and NFR (Not For Resale for reseller partners).

License Type

NVIDIA typically licenses enterprise software per GPU. Other models include per Concurrent User (CCU) and Usage-based (e.g., per GPU hour).

Part Number

The Enterprise Product Part Number that is ordered from NVIDIA channel partners defines the subscription duration, License Types and other subscription attributes for Enterprise Products.


5. NVIDIA Enterprise Support Policy

Support for Enterprise Products is governed by a unified NVIDIA Enterprise Support Policy, which is an integral part of the license agreement.

  • Business Standard (Included): Provided with all subscriptions, offering 8x5 local business hour access and standard response times.

  • Business Critical (Add-on): An optional 24x7 support level for mission-critical deployments, offering faster response times (e.g., 1 hour for Severity 1 issues).

6. Model Licensing & NIMs

NVIDIA Inference Microservices (NIMs) are licensed as a runtime under NVIDIA AI Enterprise. Importantly, the Models themselves (e.g. Nemotron and community models) are licensed separately from the NIM container.

  • Model Licenses: Models may be licensed under an Open Source license (e.g. OpenMDW or Apache 2.0), the NVIDIA Open Model License, the NVIDIA Open Model Agreement, the NVIDIA Community Model License or other model licenses. Always check the Model Card on NGC for specific terms.

7. Embedded Licensing for Partners

The Embedded License Model allows partners to integrate and redistribute NVIDIA software into their own products or services. Under this model:

  • The Partner is the primary licensee and provides Frontline Support to end-users.

  • The End-User accepts the partner's EULA rather than NVIDIA's.

  • NVIDIA provides Backline Support to the partner for issues specific to NVIDIA software components.

8. Purchasing via Channel Partners and Resellers

Orders for NVIDIA software subscriptions are placed through our extensive channel partner network, rather than directly with NVIDIA.

  • Where To Purchase: Subscriptions can be purchased through NVIDIA Partner Network (NPN) resellers, Original Equipment Manufacturers (OEMs), and Cloud Service Provider (CSP) marketplaces. Locate an NPN reseller via the NVIDIA Partner Network Locator.

  • Purchase Agreements: When buying from an NPN reseller, OEM or a CSP marketplace, the purchase agreement (covering price, quantities, and payment terms) is strictly between the you and the channel partner.

  • Unified Licensing: Regardless of the purchase route, the use of the software is always governed by the NVIDIA Software License Agreement between NVIDIA and the end customer.

Frequently Asked Questions (FAQ)

General Licensing & Agreement Structure

  • NVIDIA uses the NVIDIA Software License Agreement, which serves as the primary vehicle for enterprise software licensing.
  • The NVIDIA License Agreement covers most NVIDIA software products offered under a paid subscription with support, including NVIDIA AI Enterprise (NVAIE), NVIDIA Mission Control, NVIDIA Confidential Computing and NVIDIA networking software.
  • This means that the same terms apply across all NVIDIA software offered under a subscription.

  • Product Specific Terms are addenda to the NVIDIA Software License Agreement that define terms for individual software products.
  • For example, Networking software has its own Product Specific Terms within the overarching NVIDIA Software License Agreement.
  • These terms cover specific license grants, product definitions, and license restrictions unique to those offerings.

  • CUDA, cuDNN, and similar developer free SDKs are licensed under separate SDK agreements, which are distinct from the NVIDIA Software License Agreement.
  • These SDK agreements are available through the NVIDIA Documentation Hub
  • But when included and packaged by NIVIDIA in a container available on NGC, the NVIDIA Software License Agreement and applicable Product Terms govern use of all NVIDIA software in the container, including CUDA, cuDNN and other NVIDIA libraries.

  • NVIDIA software may include or be distributed with OSS (i.e. Separate Components), which are subject to applicable OSS Licenses or other third-party license terms.
  • Depending on the open-source license, NVIDIA passes those licenses through to the licensee.
  • To provide transparency, NVIDIA publishes Software Bill of Materials (SBOMs) in JSON format on the NGC catalog for NVIDIA AI Enterprise containers, which list all packages and security scanning details.

Pricing & Purchasing Models

  • Annual subscription: Pricing based per-seat or per-GPU-year.
  • Consumption / Pay-as-you-go (PAYG): Available for CSP marketplaces, usage-based pricing, typically per GPU hour.
  • Concurrent User Licensing (CCUs): Available for vGPU products, a license that allows concurrent Authorized Users to use the Software, up to a maximum simultaneous number.

  • The NVIDIA Software License Agreement runs for the duration of your software subscription or use of NVIDIA software.
  • Subscription terms are typically 1, 3 or 5 years.

  • Yes, most of NVIDIA’s software is available for free, including under open-source licenses.
  • Many free AI products do not require a subscription for use or distribution.

AI Models & NIM Licensing

  • No, the NVIDIA Software License Agreement licenses the NVIDIA Inference Microservices (NIMs) runtime (the container), not the underlying AI model.
  • The NIM base container is licensed under the NVIDIA Software License Agreement.
  • NIMs use both NVIDIA and third-party AI models, each of which is subject to separate licensing.

Security, Data Privacy & Compliance

  • NVIDIA enterprise software is designed for on-premises use (or in customer-controlled cloud instances), so NVIDIA does not host or manage customer data. Therefore, many standard vendor security requirements do not apply to the enterprise software.
  • To handle security due diligence, NVIDIA established the "Customer Trust" initiative to provide fast, reliable access to security, privacy, and compliance information.
  • Customers and partners can access the NVIDIA AI Trust Center for artifacts like ISO certifications and SIGLITE reports.

  • No, data collection does not include customer or Customer End User prompts, inputs, outputs, or other data or content submitted through or processed using the Software.
  • Such data and content remain within Customer-controlled or End Customer-controlled environments and are not transmitted to NVIDIA.

Support

  • NVIDIA Enterprise Support provides technical support for licensed NVIDIA software as documented in the NVIDIA Enterprise Support Policy.
  • The support model via partners involves Frontline Support (provided by the partner, reseller, or cloud provider directly to end customers) and Backline Support (provided by NVIDIA to the partner).

  • Business Standard Support is included in NVIDIA software subscriptions and provides access to telephone technical assistance during local business hours (8 AM - 5 PM on NVIDIA's working business days).
  • Business Critical Support is available as an add-on (typically for a 25% uplift) and provides 24x7x365 remote technical support for severity one (S1) and severity two (S2) issues.
  • Business Critical Support also features accelerated initial response times, such as a 1-hour response time for S1 issues, compared to 4 hours under Business Standard Support.

Specific Product Licensing

  • Confidential Computing (CC) is a security capability that protects data-in-use on NVIDIA GPUs.
  • It requires a separate license under the NVIDIA Software License Agreement.

Partner & Embedded Licensing

  • Reseller Model: The end-customer is the licensee and receives the NVIDIA software entitlement directly from NVIDIA. NVIDIA provides Frontline Support to the end-customer.
  • Embedded Licensing: The partner is the licensee and receives the entitlement. The partner provides Frontline Support to their end-customers, while NVIDIA provides Backline Support to the partner (see the Support section above for definitions).