Nvidia teams with VMware for generative AI customization

Nvidia and VMware announced at VMware Explore 2023 this week a major expansion of their ongoing partnership and a new platform aimed at helping enterprises--namely the hundreds of thousands using VMware cloud services--customize generative AI technology for their own business purposes.

Customization of generative AI has been a growing theme for Nvidia as the technology has been taking over the world in the last 10 months or so. It believes that customization allows enterprises to harness the potential power of the technology without leaving them exposed to security and privacy risks.

When it is released early next year, the new platform, VMware Private AI Foundation with NVIDIA, will do that by allowing enterprises to leverage Nvidia accelerated computing and generative Ai software to customize models and run generative AI applications, including intelligent chatbots, assistants, search and summarization. 

Nvidia and Vmware are acknowledging that the enterprise data to be used to feed and customize generative AI is located in many clouds and locations across enterprise-wide data center infrastructures. They want to give generative AI access to all this data,, but they also want to make sure they do so in a secure and cost-effective way, according to Paul Turner, vice president of product management, vSphere, at VMware.

He explained, “If data is your asset, then we have to look at how do we manage, protect, secure that data? How do you make sure that you've got confidence in it? One of the things that we believe companies will do is they'll bring more of the Gen AI workloads to their data versus moving their data into public cloud services.”

Turner said that partners believe any enterprises may want to do this “in relatively small environments,” such as a cluster of two to four, or possibly eight GPU-based server environments. Although, as enterprises are looking for greater performance as they work on generative AI projects, the platform can leverage Nvidia’s storage and NVLink and NVSwitch capabilities integrated with VMware vSphere to create configurations of up to 16 GPUs. “Then, you've got an accelerated path as you're feeding data into these models,” Turner said.

These can be pooled resources that can be shared across departments, which will help keep costs under control, he said, adding that privacy and security will be preserved by enabling customers to easily run AI services adjacent to wherever they have data. The platform will feature Nvidia NeMo, the company’s end-to-end, cloud-native framework included in Nvidia AI Enterprise — the operating system of the Nvidia AI platform — that allows enterprises to build, customize and deploy generative AI models virtually anywhere. NeMo combines customization frameworks, guardrail toolkits, data curation tools and pre-trained models.

Justin Boitano, vice president of enterprise AI at Nvidia, added that the time for enterprise to start customizing generative AI is now, as an explosive technology trend continues to reach into every industry.

“Consulting firms like McKinsey see generative AI really as having the potential to create business value equivalent to an additional $4.4 trillion a year, which is pretty amazing,” he said during a press and analyst briefing. “Generative AI is going to be used across industries to drive faster ideation, enabling designers to generate many iterations of a design concept, to help folks automate complex and redundant tasks for us… It can help us boost our creativity, discovering new ways to communicate across many different modalities using text or images or 3C objects… It's going to help us explore new areas, quickly learn new skills, new techniques. I think of these as assistants… infused in all aspects of businesses in every sector of every industry, and it’s going to deliver a ton of untapped potential.”