Dell Bets Enterprises Will Choose On-Premises Agentic AI Over the Cloud

Running AI agents in the cloud is expensive — and Dell is betting that enterprises have noticed.

At Dell Technologies World in Las Vegas, the company unveiled Dell Deskside Agentic AI, a new system built on Nvidia's NemoClaw software stack that lets organizations develop and deploy AI agents on their own hardware. Dell's pitch is direct: organizations can break even versus public cloud API costs in about three months and cut spending by up to 87% over two years.

The context matters. Agentic AI systems — designed to complete complex, multi-step tasks without human intervention — run continuously and consume inference tokens at a rate that can make cloud costs escalate fast. Dell cited one developer who generated a $3,400 bill after exhausting a billion tokens in a single day.

Key Takeaways

• The Deskside Agentic AI system pairs Dell workstations with Nvidia NemoClaw, scaling from lighter Pro Max units to full towers handling models up to 1 trillion parameters.
• Dell's AI Factory has now reached 5,000 customers — up 1,000 since February — with deployments at Eli Lilly, Honeywell, and Samsung.
• New ecosystem partnerships bring Google Gemini, OpenAI Codex, Palantir Foundry, and SpaceX's Grok models to Dell on-premises deployments.

Sam Grocott, SVP of product marketing at Dell, put it plainly: "Most enterprises don't have an AI ambition problem. They have an AI execution problem." On-premises infrastructure may be how they close that gap.

Read the full article on Quartz