VMware Partners with Nvidia on Generative AI for the Enterprise

VMware has joined the AI party and has partnered with Nvidia to develop a new software solution that will enable enterprises to build their own generative AI applications.

At his keynote speech at the VMware Explore 2023 conference in Las Vegas, VMware CEO Raghu Raghuram unveiled VMware Private AI Foundation with Nvidia, a complete software solution that will allow businesses to take their own proprietary data and customize their own generative AI models.

Raghuram introduced the concept of private AI, an architectural approach that balances the business gains of AI with the privacy and compliance needs of an organization. To protect data, Raghuram said enterprises want to pursue generative AI projects without having to move their data. They want to deploy their AI workloads where their data is located, whether it’s in data centers, the edge or the public cloud, he said.

VMware’s new generative AI software package, which integrates VMware Cloud Foundation with Nvidia AI Enterprise and includes new software components, allows organizations to train and tune their models where their data is, which helps with privacy and security of that data.

“You’ve got to get your compute and your AI to where the data is. This is inherently a multi-cloud problem, and so the solution has to be a multi-cloud solution,” Raghuram said, speaking on the new VMware Private AI Foundation software offering.

Related:Nvidia Touts New AI, Omniverse Tools and Hardware

Nvidia CEO Jensen Huang joined Raghuram on stage and said the two companies – and hundreds of their engineers – worked together on the new generative AI solution for several years.

He said VMware reinvented enterprise computing a quarter of a century ago with server virtualization and, together, VMware and Nvidia are reinventing enterprise computing again to enable enterprises to transition to accelerated computing and generative AI.

“Instead of virtualizing applications to run on CPUs, we virtualized the GPU,” Huang said at the keynote. “We  made it possible for VMware to run bare-metal performance with all of its security, and all of its manageability and vMotion capabilities across multiple GPUs and multiple nodes, to process these giant language models, and also to be able to deploy it end to end and to be completely multi-cloud. So for the very first time, enterprises around the world will be able to do private AI at scale – deployed in your company – and know that it’s fully secure and multi-cloud.”

Analysts say VMware Private AI Foundation with Nvidia is an important offering for VMware. The rapid growth of ChatGPT and other generative AI apps has prompted an increasing number of enterprises to begin pursuing their own generative AI initiatives so they can improve customer service, increase worker productivity, and enhance business operations.

Related:Broadcom’s $61 Billion VMware Deal Faces In-Depth EU Probe

“Most enterprises don’t want to be in the AI business. They simply want to consume and leverage AI, so they are looking for solutions that simplify and accelerate deployment,” said Gary Chen, IDC’s research director of software defined compute. “A lot of VMware’s messaging has been around their differentiation for privacy and filling the market needs for private AI, where people are concerned about IP risk, private data, and private access, and I definitely do see that as an important market segment.”

Nvidia is a natural partner for VMware not only because it makes GPUs to power AI workloads, but because Nvidia has created an entire software stack and AI ecosystem around its GPUs, which is just as important as the chips themselves, Chen added.

Broadcom’s Pending Acquisition of VMware

At the keynote, before announcing the AI software and other new VMware products, Raghuram and Broadcom CEO Hock Tan addressed Broadcom’s pending acquisition of VMware, which was first announced 15 months ago in May 2022.

On Monday, Broadcom announced that it expected to close the deal on Oct. 30 after the United Kingdom approved the deal.

“Over the last year, I spent a lot of time with Hock Tan, Broadcom’s CEO, and I know that he’s spent a lot of time with all of you, meeting hundreds of you in one-on-one and group meetings. And I can tell you today that he’s even more excited by the potential of VMware than he was last May,” Raghuram said.

Tan did make an appearance. In a prepared video shown at the keynote, Tan reiterated his belief that the proposed combination of Broadcom is pro-competitive and will deliver great value.

Tan added that regulatory review is in its advanced stages and that he’s confident the transaction will close. Once it does, he said Broadcom will invest $2 billion a year in VMware products and services, with half going to research and development and the other half focused on accelerating the deployment of VMware solutions through VMware and partners’ professional services.

During the keynote, VMware executives also highlighted VMware’s other announcements made today. Product updates include new Tanzu software for building and operating cloud-native applications, new VMware Cloud software and services for running, managing and securing multi-cloud environments, and new edge products.

More on VMware Private AI Foundation With Nvidia

VMware Private AI Foundation with Nvidia, which is expected to be available in early 2024, integrates a mix of existing and new software to make it easier for enterprises to pursue their own generative AI initiatives.

It includes VMware Cloud Foundation, a cloud management platform that combines compute, storage, networking and allows enterprises to run, manage and secure their workloads on-premises in core and edge data centers and across multiple public clouds. 

Also included is Nvidia AI Enterprise, a software suite that includes AI tools, frameworks and pre-trained models for enterprises to develop and deploy AI workloads, including the Nvidia NeMo framework for building, customizing and deploying generative AI models.

In addition, VMware Private AI Foundation comes with the new Nvidia AI workbench, a unified tool that allows enterprises to create, test and deploy AI models in GPU-powered workstations, data centers and the cloud.

VMware and Nvidia’s new generative AI solution also incorporates new software from VMware to help enterprises implement their AI projects, including a vector database and deep learning virtual machines, said Paul Turner, vice president of VMware’s cloud platform in a media briefing.  

VMWare executives believe many enterprises will run generative AI workloads in small 2, 4, 6 or 8 GPU-based server environments, but it can scale up to 16 GPUs.

To enable that, VMware and Nvidia today announced that Dell Technologies, Hewlett Packard Enterprise and Lenovo will build servers for VMware Private AI Foundation with Nvidia built in.  They’ll be powered by the new Nvidia L40S GPU, which provides 1.2 times more generative AI inference performance and 1.7 times more training performance when compared with the Nvidia A100 Tensor Core GPU, said Justin Boitano, Nvidia’s vice president of enterprise computing in a media briefing.

Additionally, VMware released a detailed reference architecture that shows how enterprises can deploy and optimize their AI projects with support for open-source software.

“It really takes away the complexity of how do I get started with this? Where do I go? What do I need? All of it comes prepackaged from both of us,” Turner said of the generative AI solution.