Published on March 18, 2024, 8:31 pm

A diverse range of generative AI-focused tools for developers has been unveiled in Nvidia AI Enterprise 5.0. This new version of Nvidia’s enterprise-centric AI software platform introduces a variety of microservices aimed at accelerating app development and facilitating rapid deployment, as announced today at the GPU Technology Conference.

Nvidia revealed that these microservices come in the form of downloadable software containers utilized for deploying enterprise applications. They are categorized into two main groups—Nvidia NIM, which encompasses microservices related to deploying production AI models, and CUDA-X, focusing on microservices such as cuOpt, Nvidia’s optimization engine.

The NIM microservices prioritize deployment times for generative AI apps, claiming to reduce deployment times from weeks to mere minutes with their services. Included in these microservices are Triton Inference Server for standardizing AI model deployment and TensorRT-LLM for optimizing and defining large language models, streamlining experimentation without requiring deep knowledge of C++ or Nvidia CUDA. Accessibility is extended through platforms like Amazon SageMaker, Google Kubernetes Engine, and Microsoft Azure AI, while offering support for integrations with AI frameworks such as Deepset, LangChain, and LlamaIndex.

In contrast, CUDA-X microservices focus on data preparation, model training, and tools enabling developers to link their generative AI apps to business data whether numerical information, text, or images are involved. Additionally, applications like Nvidia Riva for translation and speech AI are part of this category alongside cuOpt for process optimization and Earth-2 for climate and weather simulations.

Moreover, a plethora of new integrations is set to arrive in AI Enterprise 5.0. Businesses can utilize data from platforms such as Box, Cloudera, Cohesity, Datastax within their AI applications using version 5.0. Established hardware supported by Nvidia can be found in servers and PCs from major vendors including Dell, HPE, Lenovo among others.

Nvidia positions these microservices as a fresh layer within its comprehensive computing platform that links model developers with platform providers and enterprises while furnishing a uniform pathway for running custom AI models across clouds, data centers workstations,and PCs.

Developers can now access Nvidia’s AI Enterprise 5.0 free of charge for experimentation purposes while enterprise licenses can be procured at $4.500 per GPU per year or $1 per GPU per hour when utilizing cloud services.

In conclusion,
Jon Gold specializes in reporting on IoT and wireless networking areas for Network World but can be contacted via email at


Comments are closed.