Dell launches deskside agentic AI for local workloads
Tue, 19th May 2026 (Today)
Dell has launched Dell Deskside Agentic AI, extending the Dell AI Factory with NVIDIA.
The offering targets workgroups that want to run and scale agentic AI workflows on local systems rather than rely solely on cloud services. Dell says this approach is intended to address concerns around cost, latency and data sovereignty as companies move more AI workloads into production.
Agentic AI refers to systems that can carry out multi-step tasks with limited human intervention. Dell is positioning the deskside model for organisations that want to keep inferencing close to their data while using a common framework that can also extend into larger server deployments.
The launch adds another layer to Dell's broader AI infrastructure partnership with NVIDIA. NVIDIA OpenShell is now supported across the Dell AI Factory with NVIDIA, providing customers with a sandboxed runtime to build, test and govern AI agents from workstations through to Dell PowerEdge XE servers.
Dell is also adding support for the NVIDIA AI-Q 2.0 blueprint for multi-agent workflows. That support is now available as the Dell-NVIDIA AI-Q 2.0 Reference Architecture, built on the Dell AI Data Platform for on-premises use cases in sectors including financial services, the public sector and manufacturing.
Deskside range
The deskside package combines Dell workstations, the NVIDIA NemoClaw software stack and Dell Services. The systems are designed to handle models ranging from 30 billion to 1 trillion parameters, depending on the hardware configuration.
Three hardware options sit at the centre of the launch. The Dell Pro Max with GB10 is aimed at smaller-scale prototyping and supports models from 30 billion to 200 billion parameters. The Dell Pro Precision 9 workstation tower uses Intel Xeon 600 processors and can be configured with up to five NVIDIA RTX PRO Blackwell Workstation Edition GPUs. Dell says it supports models from 30 billion to 500 billion parameters.
At the top end, the Dell Pro Max with GB300 uses the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. It is intended for inference on larger models ranging from 120 billion to 1 trillion parameters.
The software layer includes the NVIDIA NemoClaw reference stack, which Dell describes as an open-source base for managing always-on AI agents on local hardware. The stack includes open models for reasoning and coding, alongside OpenShell's runtime environment for long-running agent workflows.
Dell Services forms the third part of the package, covering strategy, hardware deployment, workflow alignment, agent prioritisation and ongoing optimisation for customers that need support as they move AI systems into production.
Cost argument
A central part of Dell's case for local deployment is economics. The company cited analysis validated by Signal65 and Futurum Group that found organisations could break even against public cloud API costs in as little as three months for some agentic AI workloads.
Based on the same analysis, Dell said customers could reduce spending by as much as 87% over two years compared with cloud APIs. The figures were based on assumptions around multi-year deployments, a mix of workstation and server products, and workloads such as general knowledge tasks, sales support and software development.
Dell argues that rising token use in agentic AI changes the economics of cloud-heavy approaches. As tasks become more complex and involve multiple steps or agents, inferencing demand can rise quickly even if token prices are falling.
Enterprise push
The launch reflects a broader shift in the AI market as suppliers try to move customers from experimentation to operational deployment. Vendors have increasingly focused on security controls, governance and predictable infrastructure costs as barriers to wider adoption, especially in regulated industries.
Dell said more than half of agentic workflows run on open-weight models and has structured the deskside offer around that part of the market. Its argument is that many of these models can run effectively on local systems for day-to-day reasoning and operational tasks before workloads scale into the data centre.
Jeff Clarke, Chief Operating Officer, Dell, said Dell sees local deployment as a practical fit for enterprise data that remains outside public cloud environments.
"The most efficient token is the one produced closest to the data, and most enterprise data isn't in the cloud. Dell Deskside Agentic AI gives every workgroup a secure local environment to run agents, keep costs predictable and keep IP inside the building. What works at the desk scales to the data centre. That's a deployment model for the next decade," said Clarke.
NVIDIA framed the announcement as part of a broader effort to give customers one framework across desktop and server infrastructure.
"As enterprises reshape and scale the future of work with agentic AI, they're seeking infrastructure that spans the full enterprise - from our desks where work happens to the AI factories where intelligence scales. With NVIDIA OpenShell across the Dell AI Factory with NVIDIA, enterprises can develop locally, scale securely and deploy agentic AI on one consistent platform," said Boitano.
Ryan Shrout, President, Signal65, said: "As enterprises transition from AI experimentation to full-scale production, the need for secure, scalable and cost-effective infrastructure has never been greater. Dell Deskside Agentic AI, as part of the Dell AI Factory, bridges the gap between local control and enterprise scalability, while providing headroom for continued iteration. This deskside to data centre capability is unique and empowers organisations to harness the full potential of agentic AI while maintaining data sovereignty and predictable costs."