How VMware is supporting AI workloads

VMware’s Private AI Foundation, announced in partnership with Nvidia last year, is a key part of its strategy to deliver a private cloud experience for artificial intelligence (AI) workloads.

The platform’s ability to automate the delivery of AI workstations and application stacks in a matter of minutes has been a significant value proposition for customers, according to Chris Wolf, VMware’s global head of AI. “You don’t see that with a lot of the homegrown or bare metal solutions today.”

In a recent interview with Computer Weekly, he highlighted the strong traction Private AI Foundation has gained in the Asia-Pacific region, which is now second only to the US in terms of customer deployments. This momentum spans key markets such as Japan, Korea, Australia and India.

Wolf said one of the key drivers for the platform’s success has been its distributed resource management capabilities, which he described as an “almost 20-year-old technology” and a significant competitive advantage over rivals.

This has resonated with organisations struggling to manage AI resources at scale, said Wolf, noting that the platform’s ability to align graphics processing units (GPUs), memory and network resources has proven to be invaluable.

Wolf also delved into the various generative AI use cases being adopted by VMware customers, ranging from contact centres and document summarisation to coding AI assistants and financial services applications.

Notably, the ability to localise data for performance and cost benefits has seen strong traction in the retail industry, where some customers have been able to avoid outages by running AI workloads closer to the edge.

In the public sector, Private AI Foundation has also enabled government agencies to address the critical requirement of maintaining both privacy and control of data. “Our approach really resonates, because we give them the ability to run the model adjacent to their data and not have to make any changes to it,” said Wolf.

While acknowledging the value of private AI, he emphasised the importance of a hybrid approach, where organisations can develop in public clouds but deploy runtime workloads in their own datacentres to benefit from lower costs and greater control.

“It’s not all or nothing,” said Wolf. “There’s a lot beyond privacy, and a whole lot of economic value running the stack on infrastructure you control.”

He also touched on the latest developments in VMware Tanzu, focused on delivering AI capabilities to Java developers, as well as the latest features in VMware Cloud Foundation (VCF) 9 that ensure data governance and GPU availability, and reduce the cost of private AI deployments.

“We see a lot of enthusiasm from our customers around [VCF 9’s] unified policy management, unified tagging, a common object model and an easy consumption layer – these are things that they have been asking for,” said Wolf. “It’s a transformational platform for us, and it’s going to put some pressure on the industry around simplicity and how easily you can make datacentre operations.”

In a blog post, Scott Sinclair, practice director at TechTarget’s Enterprise Strategy Group, noted that the improvements in VCF 9 “aligns with the core strategic direction VMware and Broadcom are committed to moving forward – namely, delivering consolidated releases while also providing enhancements across a portfolio of capabilities that are validated and delivered as a full-stack private cloud”.

However, he added that it will be VMware’s roadmap of innovation that will play a major role in determining its long-term success amid “abrupt changes to VMware’s licensing model earlier this year led to frustration among partners and customers who experienced increased costs when switching to the new subscription model”.