As enterprises grapple with cloud costs, security worries and skills gaps, large vendors are building pre-packaged stacks to support self-hosted AI infrastructure that can deploy faster than the DIY method.
After Broadcom's launch of Tanzu Data Intelligence and expanded partnership with Canonical in August shored up its Private AI infrastructure offerings against IBM's Red Hat AI and Dell's AI Factory platforms, those companies struck back this month with updates of their own.
Red Hat provides the infrastructure components that support parent company IBM's AI orchestration and data management stack. This stack competes directly with VMware's Private AI, but emphasizes hybrid cloud deployment. Version 3 of the Red Hat AI product suite made the open-source-based llm-d AI inference utility generally available on Oct. 14 with expanded support for large language models and GPU hardware.
OpenShift AI 3.0, a component of Red Hat AI, introduced a centralized model registry and GPU-as-a-service features that make more efficient use of AI infrastructure hardware. OpenShift 4.20, launched this week, countered Broadcom's support for Canonical containers and refreshed vSphere Kubernetes Service with feature parity with the vSphere hypervisor in the OpenShift Virtualization platform.
Meanwhile, Dell's AI Data Platform and embrace of disaggregated IT infrastructure systems are seen by analysts as a successor to the hyperconverged VxRail infrastructure it previously offered when it owned VMware. There is also some overlap with Red Hat OpenShift, which Dell sells as part of an on-premises appliance for private cloud.
"To do AI, you really need to be able to scale up and scale out, not just scale out," said Rob Strechay, an analyst at TheCube Research. Dell still supports VxRail customers, but due to hyperconverged infrastructure becoming passe for AI and customer discontent with Broadcom, has made clear its strategy going forward will prioritize partnerships with Nutanix, Nvidia and other partners, according to Strechay.
Against that backdrop, Dell expanded its AI Data Platform this week with updated hardware support for its PowerScale and ObjectScale storage products. It also added preinstalled software from Elastic for vector search, Nvidia for GPU acceleration, the Starburst data analytics engine and anagentic automation layer that includes a new Model Context Protocol server.
Chasing the enterprise AI infrastructure wave
Analysts said all these vendors, along with HPE, are targeting large enterprise customers that plan to manage their own AI infrastructure. This market is still taking shape, and all face stiff competition from public cloud behemoths.
"We’re projecting that AI is going to drive cloud adoption, but security, privacy and sovereign AI requirements are also driving interest in on-premises [deployments]," said Nancy Gohring, an analyst at IDC, in an email to Informa TechTarget. "Security consistently tops the list of business risk and implementation concerns related to agentic AI in my research. I’m also finding that enterprises expect to build more custom agents in the future, which may drive interest in on-premises infrastructure for security reasons."
Respondents to IDC's Global AI Tech Buyer Sentiment Survey planned to spread AI development and deployment among cloud and non-cloud environments.
However, so far, IDC survey respondents have no strong preference for deployment locations for AI projects. In IDC's June 2025 Global AI Tech Buyer Sentiment Survey, less than half of respondents planned to use public cloud environments for the development and deployment of AI assistants and AI agents, but the remaining percentage was split between hybrid cloud, on-premises and edge computing environments.
There's also little strong differentiation among the various vendors' AI infrastructure stacks, said Jack Gold, principal analyst at J.Gold Associates. There are some subtle differences in focus -- HPE, with its acquisition of Juniper, seems mostly focused on networking, while Dell plays to its traditional strengths in storage and servers.
"HPE seems to be more componentized and more focused on the networking and data access environment, whereas Dell seems to be more focused on the end result," he said. "[But] is there one [vendor] that is so superior that it's a foregone conclusion that I'm going to go down that path? Right now, I have to say the answer is no."
OpenShift Virtualization ups ante for vSphere
Overall, IT buyers will likely look toward whichever incumbent IT infrastructure vendor they work most closely with for AI support, according to Gold and Strechay.
It has to be as fast as the cloud -- go out to those vendors and say, 'Hey, what are you doing to make this easy for me?'
Rob StrechayAnalyst, TheCube Research
"My advice [for IT leaders] would be to look at your current stack, and look for where your best negotiating power is," Strechay said. "It has to be as fast as the cloud -- go out to those vendors and say, 'Hey, what are you doing to make this easy for me?'"
One place where enterprise IT organizations might be more likely to move away from a familiar vendor is Broadcom, which has ruffled some feathers among customers with drastic licensing and pricing changes.
Hence, vendors such as IBM and Red Hat have developed products within their AI infrastructure stacks that cater to vSphere users, such as OpenShift Virtualization. In case there was any doubt about Red Hat's desire to lure vSphere users, the company also shipped version 8 of its Migration Toolkit for Applications this week, featuring another Lightspeed assistant that can automatically migrate apps from Cloud Foundry, which underpins VMware's Tanzu platform, to OpenShift.
Red Hat has also worked to catch up with vSphere directly in the hypervisor space and make OpenShift Virtualization more user-friendly to appeal to VMware users, according to Strechay.
For example, in OpenShift version 4.20, OpenShift Virtualization now supports live migration of VMs across server clusters, including offloading migration processing to external storage arrays to move large VMs more quickly. It also offers simplified VM management and a VM administrator UI integrated directly into the OpenShift web console, supported by the OpenShift Lightspeed AI assistant and OpenShift Virtualization Operator. Finally, OpenShift Virtualization now enables deeper integration into external networks for VMs.
"Lightspeed and the new UI will help people administer OpenShift, which, compared to VMware, is a science project," Strechay said. "Another big piece of this update is new OpenShift Virtualization integration with networking -- that was kind of a sore spot for people, because VMware has a lot of mature networking as part of it, and people need certain things to work, like BGP [Border Gateway Protocol] to do dynamic routing."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.
Dig Deeper on Systems automation and orchestration
