Dell is rolling out on-premises AI agent tools using Nvidia software to lure enterprises away from the cloud

At the core of the announcement is Dell Deskside Agentic AI, which pairs Dell high-performance workstations with Nvidia's NemoClaw software stack, enabling organizations to develop and deploy AI agents on their own hardware rather than sending workloads to cloud providers. Dell said organizations can break even versus public cloud API costs in about three months and cut spending by as much as 87% over two years compared to cloud-based alternatives.

The cost dynamics of agentic AI have emerged as a growing concern for enterprise buyers, given that these systems — unlike simpler chatbots — are designed to carry out complex, multi-step tasks without human intervention. Because these agents run continuously and draw heavily on inference tokens to process each step, the costs can escalate quickly in a cloud environment. To illustrate the exposure, Jon Siegal, SVP of Dell's client solutions group, told reporters about an incident in which one developer generated a $3,400 bill after exhausting one billion tokens within a single day, according to ITPro.

Shipping now, the deskside system comes in multiple hardware tiers — from smaller Dell Pro Max units designed for lighter workloads up to full tower workstations built to handle models reaching 1 trillion parameters, according to Dell. Nvidia OpenShell, which provides a sandboxed environment for building and testing agents, has been extended to run across the full Dell AI Factory lineup, from desktop workstations through to PowerEdge server infrastructure.

The announcements came Monday at Dell Technologies World in Las Vegas. Michael Dell told Bloomberg that the AI Factory lineup — a bundle of servers, Nvidia processors, software, and accompanying services — has grown to 5,000 customers, a jump of 1,000 since the company's February earnings report. Customers include Eli Lilly $LLY, Honeywell $HON International, and Samsung Electronics.

Separately, Dell detailed a set of upgrades to its AI Data Platform, adding orchestration and search features capable of indexing billions of unstructured files, alongside SQL analytics tooling built jointly with Nvidia and Starburst Data that the company says runs up to six times quicker on Nvidia Blackwell hardware. A new ObjectScale X7700 storage appliance offers up to 45% more hard-drive capacity than the previous generation.

For infrastructure, Dell unveiled PowerRack, a pre-engineered rack-scale offering that arrives with compute, networking, storage, cooling, and management already unified, targeting AI and high-performance computing environments.

Dell also announced new ecosystem partnerships. Through a partnership with Google $GOOGL, Gemini models will reach customers via Google Distributed Cloud running on Dell hardware. OpenAI's Codex agent for software development will tie into both the Dell AI Data Platform and AI Factory. Palantir $PLTR is bringing its Foundry and AI Platform products to Dell on-premises deployments, and a new arrangement with SpaceX will make Grok models available for on-site enterprise use.

Sam Grocott, SVP of product marketing at Dell, offered a pointed diagnosis of why the company sees opportunity in this space. "Most enterprises don't have an AI ambition problem," Grocott said at a media briefing, according to SiliconAngle. "They have an AI execution problem."