Dell Technologies is rolling out a wave of new AI Factory with Nvidia offerings, including a deskside agentic AI platform designed to slash costs and boost data security, as executives tout faster paths from experimentation to production.
Dell Technologies unveiled a blizzard of agentic AI Dell AI Factory with Nvidia offerings, including a breakthrough Dell Deskside Agentic AI product aimed at making agentic AI more cost-effective and secure.
“We are introducing some real exciting innovations here that help customers accelerate agentic AI from early experimentation at deskside all the way to production in the data center,” said Dell Senior Vice President of Client Solutions Group Marketing Jon Siegal in a press conference.
The Dell AI Factory with Nvidia innovation directly addresses customer concerns around the cost and data privacy of AI agent solutions, said Siegal.
The Dell Deskside Agentic AI platform, for example, provides an 87 percent reduced spend versus public cloud spend for running, building, testing and fine-tuning agentic AI solutions.
What’s more, Dell said the break-even versus using public cloud APIs is just three months.
While enterprise IT leaders are excited about the “potential game- changing benefits of agentic AI” there is “trepidation and concerns,” particularly around controlling costs, said Siegal. “Super users are burning through tokens at such a high rate that they have sticker shock from their cloud bills,” he said. “We’ve seen this at Dell firsthand where we actually had a single developer burn through 1 billion tokens in 24 hours. That was a $3,400 cloud bill.”
Customers are also concerned that “sensitive company data could be exposed or misused in some way due to security risks, data privacy concerns, etc. and just the lack of guardrails,” Siegal said.
Dell is “answering the call” of customers with seamless agentic AI workflows across the Dell AI Factory with Nvidia, said Siegal.
Dell touts the Dell AI Factory with Nvidia as the “first and only” end-to-end enterprise AI platform that scales from the desktop to the largest data center deployments.
The Dell AI Factory with Nvidia partnership, which is just over two years old, has resulted in more than 320 releases of Dell AI Factory with Nvidia offerings, said Siegal.
“More than 160 of those releases will have landed since Dell Technologies World last year,” he said. “That is roughly one meaningful update every two days, and the pace hasn’t slowed. It has actually accelerated.”
Dell Technologies Senior Vice President of Product Marketing Sam Grocott, for his part, said Dell is aiming to reduce the “complexity and decision-making” with regard to where agents and agentic workflows should be run, whether it is on-premises, at the edge or in the public cloud.
On-premises AI adoption is becoming a bigger factor, driven by data “sovereignty, security performance and increasing cost and how you take advantage and take control of the token economics,” said Grocott. “These four areas are absolutely connected and absolutely what we are solving for [with the Dell Technologies World announcements].
“The historic easy button of just running it all in the cloud is not going to last as we go forward in terms of cost, economics, security, sovereignty, etc.,” said Grocott. “What we have really done leveraging the 5,000 [Dell AI Factory with Nvidia] customers we’ve got around the world today is we have learned from them. We have documented them. We have got a set of best practices. We have got a set of Dell AI Factory services that we go to market with to help our customers really decide essentially in a hybrid AI world where some AI workloads run in the cloud, some run on-prem, some run in the data center, some run deskside. We are really helping customers lock in on making sure they run their agentic AI workloads in the right place with the right model at the right tier.”
Grocott said Dell is making sure that customers “don’t make a simple, but very expensive, a very insecure, a very risky decision by just doing AI” in one place. “That’s not the right way going forward,” he said. “Right place, right model, right tier is our best practice that we are educating our enterprise customers on now.”
Here are the biggest Dell AI Factory with Nvidia announcements from Dell Technologies World.
Dell Deskside Agentic AI: 87 Percent Reduced Spend Versus Public Cloud
The breakthrough new Dell Desktop Agentic AI system enables customers to stand up agentic AI solutions in a secure deskside sandbox to run, build, test and fine-tune AI agents locally.
Dell said the Deskside Agentic AI platform provides 87 percent savings versus public cloud spend for running, building, testing and fine-tuning agentic AI solutions. Furthermore, the company said the break-even versus using public cloud APIs is just three months.
“This is a new approach to running AI without the need to rely on cloud infrastructure and where data itself never leaves the environment,” said Siegal. “We’re creating a simple path, if you will, to bring production-ready agentic AI to workgroups.”
Among the new reference architectures for the system are a deskside coding agent assistant for software teams; a research assistant architecture for sensitive data at universities, government labs and medical institutions; and a private assistant for organizations in highly regulated industries.
The new Dell Deskside Agentic AI system runs the Nvidia NemoClaw secure operations layer to run agents securely on local infrastructure. “It provides enterprise-grade security and governance on day one,” said Siegal.
The Dell high-performance Deskside workstations supporting the platform run from the Dell GB10 to the Dell GB300 to Dell Pro Precision Towers.
“We really give customers a lot of choice, the ability to right-size their environment and their investment and run a broad range of workhorse AI models from 30 billion parameters up to a trillion parameters,” said Siegal.
‘Seamless’ Dell AI Factory With Nvidia Agentic AI Workflow Development With Support For Nvidia AI-Q 2.0 Blueprint
Dell is upping the ante on agentic AI development with what it calls its “seamless” Dell AI Factory with Nvidia Agentic AI Workflow Development platform, including “fully integrated” support for Nvidia’s AI-Q 2.0 Blueprint for multiagent workflows.
“We are going all in on complex multiagent workflows [with Nvidia AI-Q 2.0 Blueprint support],” said Siegal. “Nvidia AI-Q 2.0 Blueprint gives teams a real proven starting point. They can pilot advanced research agents, they can iterate fast, they can move directly into repeatable, production-ready systems.”
With the new Dell agentic AI “breakthrough innovations,” enterprises now have a seamless path to scale agentic AI from the deskside to the data center while still having control, said Siegal.
The agentic AI development effort also supports Nvidia’s NemoClaw secure operations layer on the Dell Deskside system.
Finally, the Dell agentic AI workflow development platform runs Nvidia’s OpenShell, which provides organizations a “secure sandbox to build and test agents everywhere from the deskside all the way to the servers,” said Siegal.
“The net result is customers can accelerate development and also at the same time de-risk when they want to move the agents into production,” said Siegal.
Dell AI Data Platform With Nvidia Omniverse Support
Dell is aiming to eliminate one of the biggest bottlenecks to agentic AI development and deployment with the Dell AI Data Platform with support for Nvidia Omniverse.
The Omniverse support allows customers to run “digital twins and physical AI on enterprise-grade data,” said Dell Senior Vice President of Infrastructure and Telecom Marketing Varun Chhabra.
“This allows [agentic AI development] teams to work directly on large-scale datasets with the performance and reliability they expect from Dell,” he said.
The Dell AI Data Platform provides customers with a unified, vertically integrated stack as a data foundation that brings analytics and AI together in a single stack.
The Dell AI Data Platform improvements also include SQL analytics support in partnership with Nvidia and AI agent data platform provider Starburst to provide up to six times faster SQL database query performance on Nvidia Blackwell GPUs. The SQL analytics support also includes future support for Nvidia Vera GPUs.
In addition, Dell has enhanced the data orchestration capabilities of the data platform so it can “quickly index billions of unstructured files and pull them into governed pipelines, making it faster for customers to build AI-ready datasets, which results in up to 12X faster vector indexing,” said Chhabra.
The storage layer of the data platform also includes updates on the storage foundation with new Dell ObjectScale x7700 delivering ultra-dense storage for AI data lakes as well as archives, said Chhabra.
The Dell AI Data Platform starts with secure, scalable AI storage with PowerScale, ObjectScale and the Lightning File System, said Chhabra. It also includes modular data engines for analytics, streaming and search, and a data orchestration engine that “prepares, governs and routes data directly to customer models,” he said.
Chhabra said the Dell AI Data Platform improvements directly address the problem customers face in addressing the “messy data” problem that can hold back agentic AI development.
“One of the hidden bottlenecks that we find when we talk to customers is the ability to get their data strategies right,” he said. “Customers are held back because their data is often messy. It is both structured and unstructured, spread across silos and very, very hard to govern. This slows down AI efforts because it makes it hard to move from pilots to real agentic AI production.”
Dell PowerRack For Compute, Networking And Storage
Dell is teaming with Nvidia to deliver Dell PowerRack, a turnkey rack-scale offering that unites compute, networking and storage engineered as scalable units validated as a single system.
“With Dell PowerRack, customers no longer need to buy components and hope they work together,” said Chhabra. “They can deploy a fully integrated system in which the thermal design, the power management and the software for infrastructure management have all been engineered to work as one unit.”
Chhabra said the production-ready PowerRacks can be set up, connected and be running live workloads in a customer data center within six and a half hours or less after delivery. “This is extreme time to value for the most extreme, dense performance- sensitive AI and HPC applications in the industry,” he said.
Dell PowerRack For Networking: Speeding Up AI Fabric Rollouts
The Dell PowerRack For Networking, a turnkey rack prebuilt and validated with all networking, power and cooling ready to go, is aimed at speeding up AI fabric rollouts.
“The system shows up ready to go, as with all PowerRacks, there are no days or weeks of on-site assembly and wiring needed,” said Chhabra. “Networking, power, cabling, cooling, all engineered together, not bolted on as afterthoughts.”
PowerRack for Networking features 800 Tbps of switching capacity with eight new Dell PowerSwitch SN6600-LDs Ethernet switches per rack. That enables the “high-volume, low-latency, east-west traffic required by GPU-dense environments,” said Chhabra. “Instead of dealing with multiple vendors, customers get one number to call and a system that is ready to scale their AI fabric right out of the box,” he said.
Dell Exascale Storage With Support For PowerFlex Software- Defined Infrastructure
Dell is taking its Exascale storage to new heights by adding support for its PowerFlex software-defined infrastructure.
The software-defined capabilities open the door for customers to use PowerFlex for high-performance workloads.
“Now Exascale storage is going to be expanding to not just AI and HPC but also to extreme enterprise workloads,” said Chhabra. “So it is the only four-in-one storage that is built for extreme-scale HPC, AI and demanding enterprise workloads. Customers can swap storage personalities on demand. There is no replatforming required. There are no forklift upgrades. This product embodies cloud-scale agility, stronger economics and built-in investment protection as AI evolves.”
The new Exascale storage capability now supports 245-Terabyte- class drives for Dell ObjectScale—tripling the PowerRack density for S3 storage. It also features up to 6 Terabytes per second of performance per rack.
“Customers looking to run large-scale AI data lakes, vector stores and long-term models can take advantage of this,” he said. “Ultimately, Exascale is really aimed at keeping high-value GPUs fully utilized so CSPs [cloud service providers] and AI-native organizations spend less time managing their storage infrastructure and more time on shipping innovation.”
Dell PowerCool CDU C7000 Efficiently Cools Nvidia Vera Rubin NVL72
Dell is releasing its most compact and at the same time robust in-rack cooling distribution unit with the release of the Dell CDU C7000.
The compact C7000—which eliminates the need for floor-mount CDUs—is aimed at cooling even the most-high-performance AI environments including the Nvidia Vera Rubin NVL 72 rack-scale AI GPUs.
The 4U rackspace system delivers more than 220 kilowatts of cooling capacity, supporting warmer facility water temperatures of up to 40 degrees. That reduces energy costs “substantially” for customers, said Chhabra.
The C7000 features advanced manageability, redundancy and enhanced leak detection capabilities, said Chhabra.
Dell OpenManage And Integrated Rack Controller Enhancements
Dell is enhancing its OpenManage systems management platform with its Integrated Rack Controller, isolating potential power, thermal or hardware issues and shutting them down before the faulty issue spreads to other nodes on the network.
“This reduces downtime, limits hardware exposure and maintains overall stability,’ said Chhabra. “This is a substantial advancement that provides better rack-level control and visibility and avoids unnecessary disruption to healthy workloads. It also means that the rack responds in a single orchestrated system, giving operators fewer surprises and more consistent outcomes.”
The continuous monitoring with OpenManage, Dell said, detects leaks in under seven seconds and takes automated action in less than three seconds.
The OpenManage platform also provides a preconfigured list of all devices and their status in the rack. “That gives operators one reliable view of everything in the rack, which helps them deploy systems faster, find and fix problems more quickly and manage everything more efficiently,” Chhabra said.
The OpenManage platform provides remote connectivity and orchestrations on the IRC across servers, CDUs and network switches with PowerShell.
“This means that the whole rack can be controlled with a single UI acting in a coordinated policy-driven way across all of its different components,” said Chhabra.
