Kathy Gibson reports from VMware Explore – As the company that coined the concept of the software-defined data centre (SDDC), and cut through the cloud hype with the multi-cloud and the hybrid cloud, VMware is now making its mark in the artificial intelligence market with the launch of Private AI.

AI is evolving, from the predictive models of the past to the rise of the data scientist and the growth of specialised AI models like computer vision, says Chris Wolf, vice-president: VMware AI Labs.

Now, large language models (LLMs) are the new trend, making an impact on any number of use cases. Indeed, generative AI (GenAI) is expected to drive $4,4-trillion in annual economic value.

We are seeing use cases across the board, in different segments, and for different uses, says Wolf. “It is everywhere – and the opportunities are immense.”

VMware started working with GenAI, using it for AI-assisted code development, and overcoming the many challenges involved led it to setting up the AI Labs to help customers through the same journey.

“We realised our use cases are the same as our customers’ use cases,” Wolf says.

At the end of the day, data plus compute represents AI, with AI itself being a multi-cloud use case.

But the core enterprise challenges around privacy are exacerbated with GenAI, with private IP, private data and private access all coming under the spotlight.

This is why VMware is driving the concept of private AI, to balance business use with privacy and the compliance needs of the organisation.

Private AI isn’t just about privacy, Wolf adds: it offers additional value in the form of choice, cost, performance and compliance.

Privacy must be at the centre of AI, but so must choice. “We cannot bet on a single stack,” Wolf says. “So you have to have a platform where optionality is built in.

“Cost is at the centre of this as well. Our internal stack shows that it comes in at one-third of the public cloud.

“Performance is also key: it needs to be faster than bare metal.

“And we need to ensure we are meeting regulatory and compliance needs.”

Earlier this year, VMware launched Private AI Foundation with Nvidia, bringing together the best of VMware with the best of Nvidia, and partnering with Dell, HPE and Lenovo.

This will be available soon, but customers can start developing now thanks the reference architecture for VMware Private AI that allows customers to run across any cloud.

VMware Private AI Open Ecosystems are enabling the development of new applications, while more global systems integrators are also joining the ecosystem.

Today, VMware is also announcing a collaboration with Intel to extend the companies’ more than two decades of innovation to help customers accelerate the adoption of artificial intelligence (AI) and enable private AI everywhere – across data centres, public clouds, and edge environments.

VMware and Intel are working to deliver a jointly validated AI stack that will enable customers to use their existing general-purpose VMware and Intel infrastructure and open source software to simplify building and deploying AI models.

The combination of VMware Cloud Foundation and Intel’s AI software suite, Intel Xeon processors with built-in AI accelerators, and Intel Max Series GPUs, will deliver a validated and benchmarked AI stack for data preparation, model training, fine-tuning and inferencing to accelerate scientific discovery and enrich business and consumer services.

“For decades, Intel and VMware have delivered next-generation data center-to-cloud capabilities that enable customers to move faster, innovate more, and operate efficiently,” says Sandra Rivera, executive vice-president and GM of the Data Centre and AI Group (DCAI) at Intel. “With the potential of artificial intelligence to unlock powerful new possibilities and improve the life of every person on the planet, Intel and VMware are well equipped to lead enterprises into this new era of AI, powered by silicon and software.”

VMware Private AI brings compute capacity and AI models to where enterprise data is created, processed, and consumed, whether in a public cloud, enterprise data center, or at the edge, in support of traditional AI/ML workloads and generative AI.

VMware and Intel are enabling the fine-tuning of task specific models in minutes to hours and the inferencing of large language models at faster than human communication using the customer’s private corporate data. VMware and Intel now make it possible to fine-tune smaller, economical state of the art models which are easier to update and maintain on shared virtual systems, which can then be delivered back to the IT resource pool when the batch AI jobs are complete. Use cases such as AI-assisted code generation, experiential customer service centers recommendation systems, and classical machine statistical analytics can now be co-located on the same general purpose servers running the application.

VMware and Intel are designing a reference architecture that combines Intel’s AI software suite, Intel Xeon processors, and Data Centre GPUs with VMware Cloud Foundation to enable customers to build and deploy private AI models on the infrastructure they have, thereby reducing total cost of ownership and addressing concerns of environmental sustainability.

This VMware Private AI reference architecture with Intel AI will include:

  • 4th Gen Intel Xeon processors with Intel Advanced Matrix Extensions (Intel AMX) deliver up to 10x significant out-of-box performance improvements using standard industry frameworks and libraries, end-to-end data science productivity tools, and optimised AI models.
  • Intel Data Centre GPU Max contains up to 128 Xe cores and is Intel’s foundational GPU compute building block targeted at the most demanding AI workloads. Intel Max Series GPUs will be available in several form factors Ito address different customer needs.
  • Intel’s AI software suite is packaged with end-to-end open source software and optional licensing components to enable developers to run full AI pipeline workflows from data preparation to fine-tuning to inference, accelerate building multi-node scaling and deploying AI on enterprise IT infrastructure. The open oneAPI framework enables processors and hardware accelerator-agnostic software development, allowing developers to write code once and run it across different architectures, eliminating the need for multiple code bases and specialized languages. Intel’s Transformer Extensions and PyTorch Extension’s deep integration with developer favorite open source Hugging Face libraries provide automated optimization recipes for fine-tuning and compressing models for efficient inference.

VMware Cloud Foundation brings consistent enterprise-class infrastructure, operational simplicity, and enhanced security to VMware Private AI through capabilities such as:

  • VMware vSAN Express Storage Architecture which offers up to 30% greater performance1 and minimal overhead for important capabilities such as encryption for I/O intensive AI/ML tuning and inferencing workloads.
  • vSphere Distributed Resources Scheduler helps efficiently use spare compute capacity for AI model training and inferencing, leveraging the same clusters other AI applications use, thereby maximising capacity and reducing TCO.
  • VMware NSX for building micro segmentation and advanced threat protection capabilities.
  • Secure boot and Virtual TPM to enable model and data confidentiality.