Dell Technologies World 2026 in Las Vegas spotlighted a major shift in how the tech industry plans to deploy AI, with Dell and NVIDIA outlining a strategy focused on local AI processing and agentic systems.

A common theme ran throughout the event: enterprises want AI systems that operate closer to users, data and workflows rather than relying mostly on centralized cloud platforms.

Dell CEO Michael Dell and NVIDIA CEO Jensen Huang described the transition as the next evolution of computing architecture. Speaking during a Bloomberg Television interview, Huang described the emergence of “personal AI,” saying that AI systems must function where “context” exists. That is, relevant data on laptops, at the network edge and within autonomous systems.

Dell promoted the PC as critical for enterprise productivity, particularly as companies deploy devices capable of running smaller AI models locally. The company is embedding hybrid AI capabilities into workstation and PC platforms to support on-device inferencing and autonomous AI applications.

“In the next 2-3 years we expect a major shift in how companies utilize AI,” said Jack Gold, Principal Analyst at J. Gold Research.

“Currently the vast majority of AI is done with cloud-based compute at hyperscalers or large data centers. While we don’t expect this strategy to go away entirely, we do expect a much broader range of AI compute resources to be deployed, particularly as inference and agentic AI take hold in the next 1-2 years, and AI models move from predominantly LLMs to focused SLMs that are more task-specific.” In Gold’s view, Dell is well positioned for this transition.

Agentic AI and NemoClaw

Huang explained that LLMs now function as the “brain” of agentic AI systems, while additional software layers allow the models to interact with enterprise tools and networks.

To support these workloads, Dell expanded its Dell AI Factory with NVIDIA initiative, which the company said now supports more than 5,000 enterprise customers. The updated portfolio includes new hardware, cooling systems and AI software integrations, all with a goal to move enterprises from AI pilots into production deployments.

Dell debuted Deskside Agentic AI, a local AI platform combining Dell workstations, NVIDIA NemoClaw software and Dell services. Deskside enables organizations to build and run AI agents directly inside their own environments instead of relying exclusively on cloud-hosted AI services.

The value here is cost control and protecting sensitive data. On the cost issue, Dell notes that agentic AI workloads can run up unexpectedly high cloud invoices because agents continuously consume inference tokens while executing workflows.

The Deskside systems range from compact Dell Pro Max configurations to large workstation towers built for demanding local AI workloads.

Dell also announced support for NVIDIA OpenShell, a runtime environment for developing, governing and fine-tuning autonomous agents locally across Dell AI infrastructure.

Orchestration Tools

Dell executives repeatedly stressed that many enterprise AI projects stall because companies struggle to govern and organize disparate and siloed data environments.

To address that challenge, Dell introduced updates to the Dell AI Data Platform, including orchestration tools that can index billions of unstructured files. The company also unveiled GPU-accelerated SQL analytics developed with NVIDIA and Starburst, which Dell claimed can improve query performance by up to six times on NVIDIA Blackwell GPUs.

Dell expanded integration with NVIDIA Omniverse, the simulation and digital twin platform used for robotics and physical AI applications. The integration allows enterprises to connect storage systems, semantic search and digital twin workflows inside AI development environments.

Infrastructure announcements at the conference demonstrated growing demand for turnkey AI deployments. Dell introduced PowerRack, a pre-engineered rack-scale system integrating compute, networking, storage, cooling and power management into unified AI infrastructure.

The company also unveiled a rack-mounted cooling distribution unit designed for NVIDIA’s future Vera Rubin NVL72 platform, highlighting how thermal management is a critical challenge as AI systems devour more energy.