Dell Pushes Disaggregation for the Private Cloud

(Editor's note: This is a Cloud Tracker Pro article that will be available for limited time for free.)

To no one's surprise, AI was the star of Dell Technologies World, the global conference being held by the IT giant this week in Las Vegas. The May 19 opening keynote featured a blitz of announcements around the AI factory, including the introduction of the Dell AI Factory with Intel -- the matching piece next to Dell's AI work with NVIDIA and AMD. But Dell's more strategic side got the spotlight on Day Two (May 20), as the company explained its vision for the disaggregated datacenter.

The goal is flexibility: having pools of storage and computing resources that workloads can tap as needed. Hyperconverged infrastructure preached a similar message but became associated with vendor lock-in. Dell wants the concept to become more open.

Specifically, Dell's VxRail HCI products were built for VMware's vSphere. With VMware now owned by Broadcom and enterprises entertaining alternatives, it only makes sense for Dell to steer that HCI story toward more open waters, inviting alternatives -- specifically Red Hat OpenShift and Nutanix's AHV.

The endgame here isn't just a new type of HCI rack. The entire datacenter has become a product in a sense -- that's what NVIDIA's AI Factory concept is all about -- and Dell wants to be the one to place that product into the enterprise, leaving space for choices of software and operating systems.

Hence, Dell announced the Dell Private Cloud today, a turnkey cloud that the enterprise purchases just as they would purchase an appliance. It supports multiple hypervisors, as noted above, and customers can bring their own cloud OS licenses as well. This isn't just an AI thing, either; Dell is championing disaggregation as a key to the future of datacenter architectures in general.

Along similar lines, Dell NativeEdge is expanding its scope. In this case, it's about more easily encompassing the enterprise's existing estate. VMware virtual machines can now be easily imported into NativeEdge. More importantly, NativeEdge capabilities such as zero-touch onboarding can now be applied to the third-party hardware a customer is already using. As with Dell Private Cloud, this is not solely targeted at AI; it's all meant to further modernize edge deployments in general.

To remove the guesswork of how to get these pieces operating together, Dell has launched the Dell Automation Platform as the single portal for managing Dell Private Cloud, NativeEdge, and certain Dell AI solutions.

Beyond disaggregated

Dell might be able to go further with disaggregation, judging from what we've seen before in the ecosystem. One possibility is the grand goal of composable datacenters, where compute, storage, and networking capabilities can all be ratcheted up or down, directed by fungible hardware.

This was the dream of the company aptly named Fungible, founded in 2015, which used the data processing unit (DPU) as the distributed workhorse of this architecture. The idea was too early and complex to catch on with enterprises, however; Fungible found use cases in storage but never really broke that orbit and eventually got acquired by Microsoft in 2023.

Fungible was right about DPUs, though, as they've become a centerpiece of cloud infrastructure, even for the AI-specific players known as Neoclouds. At a GTC session in March, Coreweave explained its orchestration platform, Nimbus, which is executed in NVIDIA BlueField DPUs. During the same session, an NVIDIA presenter noted the company was bringing the Kubernetes control plane into BlueField DPUs.

The point is that there is further to go with disaggregation, and the DPU might be the crux of it all. Dell integrates NVIDIA's DPUs with PowerEdge servers and soon with Dell ObjectScale storage (as announced this week, and due for availability in 2H'25). It hasn't had to move more aggressively into a DPU-driven world, but it might get pulled that way. Regardless, the private cloud is now an enterprise unit of consumption and should be a corner of Dell worth watching.

Enter the DPUs

On a more down-to-earth level, Dell did announce plenty of hardware and software related to its AI Factories and AI Platforms, including new servers built around NVIDIA, AMD, and Intel.

That included launching the Dell AI Platform with Intel, completing the trifecta of AI platforms for the three big processor vendors. The Dell-Intel platform, now shipping, is built around Gaudi 3 AI accelerators and Xeon processors.

The corresponding Dell AI Platform with AMD was introduced last summer. This week marked the fourth set of upgrades to the platform, including using the AMD ROCm 6.3.1 stack and pushing storage bandwidth to 200 Gb/s. And of course, Dell showed plenty of love for the Dell AI Factory with NVIDIA. Announcements there included support for the NVIDIA GB300 NVL72 in the rack-scale Dell PowerEdge XE9712 and the integration of NVIDIA's Quantum 3 InfiniBand into PowerSwitch SN5600 switches.

Futuriom Take: Even with NVIDIA taking over the AI stack, Dell and its peers have a role to play as the conduit to real-world large enterprises. That's especially true for the enterprises that want AI deployments outside the cloud. The turnkey private cloud is the right direction for Dell, as is the message of disaggregation.