NVIDIA announced on the 16th that it is providing AI Foundry services to Microsoft Azure. This will provide businesses, including startups, with enhanced development and tuning of custom, generative AI applications.

NVIDIA AI Foundry Service integrates NVIDIA AI Foundation models, NVIDIA Nemo framework, and NVIDIA DGX cloud AI supercomputing service to provide an end-to-end solution for enterprises to build custom generative AI models. This allows businesses to deploy custom models alongside NVIDIA AI Enterprise software to power generative AI applications that support intelligent search, summarization, content creation, and more.

Industry leaders SAP SE, Amdocs, and Getty Images are using the service to build custom models.

“Enterprises need custom models to perform specialized tasks that are trained on their data, a part of their unique DNA,” said Jensen Huang, founder and CEO of NVIDIA. “The NVIDIA AI Foundry service combines NVIDIA’s generative AI model technology, our large-scale language model (LLM) training expertise, and our large-scale AI factory.”

Nvidia said it built this on Microsoft Azure, allowing businesses around the world to connect their custom models to Microsoft's cloud services.

“Our partnership with NVIDIA spans every layer of the Copilot stack, from silicon to software, as we innovate together for the new era of AI,” said Satya Nadella, Microsoft chairman and CEO. “NVIDIA’s Generative AI Foundry service gives enterprises, including startups, new capabilities on Microsoft Azure to build and deploy cloud-native AI applications.”

■ Industry leaders building customized LLMs

NVIDIA AI Foundry services are used to customize models for generative AI-based applications across industries, including enterprise software, communications, and media. Once ready for deployment, enterprises can use Retrieval Augmented Generation (RAG) technology to connect models with enterprise data and access new insights.

SAP is the first customer for NVIDIA DGX Cloud on Microsoft Azure. SAP plans to use this service and its optimized RAG workflows in conjunction with NVIDIA DGX Cloud and NVIDIA AI Enterprise software. They run on Azure and help customize and deploy Juul, a new natural language generation AI copilot.

“JUUL leverages SAP’s unique position at the intersection of business and technology and is built on a foundation of a relevant, trusted and responsible approach to business AI,” said Christian Klein, CEO and member of the executive board of SAP SE. “Through our partnership with NVIDIA, JUUL is helping customers realize the potential of generative AI for their businesses by automating time-consuming tasks, rapidly analyzing data and delivering more intelligent and personalized experiences.”

Amdocs, a provider of software and services to communications and media companies, is optimizing models for the amdocs amAIz framework to accelerate the adoption of generative AI applications and services by telecom operators worldwide.

“By leveraging NVIDIA and Microsoft technologies to enhance the Amdocs AmazE framework, we can deliver new generative AI-powered applications to our customers more quickly,” said Shuki Schaefer, Amdocs President and CEO. “Furthermore, we will be able to leverage the tremendous potential of generative AI while providing enterprise-grade security, stability, and performance.”

■ Optimized model for custom generative AI

Customers using NVIDIA Foundry services can choose from a variety of NVIDIA AI Foundation models available in the Azure AI Model Catalog, including the new NVIDIA Nemotron-3 8B family of models. Developers can access the Nemotron-3 8B model in the NVIDIA NGC catalog. Additionally, community models such as Meta's Rama2, which is optimized for NVIDIA for accelerated computing, will soon be available in the Azure AI model catalog.

Optimized with 8 billion parameters, the Nemotron-3 8B family includes versions tailored to a variety of use cases. It also features multilingual capabilities for building custom enterprise-grade AI applications.

NVIDIA DGX Cloud AI supercomputing is now available on the Azure Marketplace. It scales to thousands of NVIDIA Tensor Core GPUs via user-rentable instances. It also comes with NVIDIA AI Enterprise software, including Nemo, to accelerate LLM customization.

With the addition of DGX Cloud to the Azure Marketplace, Azure customers can accelerate model development by leveraging NVIDIA AI supercomputing and software with their existing Microsoft Azure consumption commitment credits.

The integration of NVIDIA AI Enterprise software into Azure Machine Learning adds a secure, reliable, and supportable NVIDIA AI and data science software platform. Nemo and NVIDIA Triton inference servers are now available for deployment as part of Azure's enterprise-grade AI services.