Dell and NVIDIA scale agentic AI with Pro Precision and PowerEdge

Dell Technologies and NVIDIA announced a broad set of advancements to the Dell AI Factory with NVIDIA on 21 October 2026, introducing hardware for agentic AI—AI that can act on its own—ranging from workstations to data center infrastructure. The expansion includes the Dell Pro Precision with GB300, which offers 748GB of coherent memory, and several new PowerEdge models designed to handle specific AI workloads. (coherent memory: shared memory that is accessible by both the central processing unit and the graphics processing unit.) These updates aim to help enterprises move AI projects from experimentation to production on infrastructure they own and control.

If you already own a public cloud-only AI infrastructure, these hardware announcements provide a path to transition toward on-premise ownership to manage costs and data control. The move centers on providing a complete end-to-end AI-ready data foundation.

Deskside Agentic AI and the Pro Precision GB10 and GB300

Dell Deskside Agentic AI Solutions for NVIDIA provide reference templates to scale autonomous AI workflows from a single developer to over 1,000 systems. These solutions specify the hardware, models, and software stacks needed to run use cases like agentic coding and the Hermes Agent on NemoClaw. Developers can find these solutions on Github to begin local setup.

The Dell Pro Precision with GB10 provides a 64GB unified memory configuration for individual

The Dell Pro Precision with GB10 provides a 64GB unified memory configuration for individual developers and smaller teams. It becomes available on 23 October 2026. For larger requirements, the Dell Pro Precision with GB300 brings 748GB of coherent memory—shared memory accessible by both the CPU and GPU—and 20 petaflops of FP4 compute to Windows environments. The GB300 provides sufficient memory capacity for enterprises to run large open-weight models directly on Windows desktops. (open-weight models: machine learning models where the internal parameters are available for public use and local deployment.)

Data center infrastructure for the PowerEdge M9882 and XE9810

The Dell PowerEdge M9882 server is designed for CPU-based agentic AI execution, including orchestration, tool use, and code execution. It packs 256 NVIDIA Vera CPUs into a single 48U rack. By separating CPU compute from GPU clusters, organizations can deploy the NVIDIA Vera CPU Rack to maximize data center power and cooling capacity instead of over-provisioning mixed-use servers. The PowerEdge M9882 allows for the separation of CPU compute for orchestration from GPU clusters to maximize data center power and cooling.

High-throughput inference is addressed by the Dell PowerEdge XE9810, the first to deliver rack-scale inference featuring 16 direct-liquid-cooled compute trays. These trays are powered by Groq 3 LPUs—AI language chips designed for high-throughput inference. This configuration enables high-throughput inference as a disaggregated complement to existing GPU clusters. (disaggregated complement: a hardware configuration where different computing components are separated to work together rather than being integrated.) The PowerEdge R8822 also joins the lineup this month, offering air-cooled two-socket NVIDIA Vera CPU performance in a standard 19-inch rack.

Modular data centers and the DSX POD

The Dell Modular Data Center with NVIDIA DSX POD provides a containerized AI foundation. It includes compute, storage, networking, rack-level power, and liquid cooling as part of its foundation. This design allows organizations to connect to prepared power, which reduces onsite integration complexity. The modularity allows for rapid deployment of a complete AI factory in a containerized format.

Economic advantages of owned AI infrastructure

Dell claims the Dell AI Factory can be up to 90% more cost-effective than public clouds over two years. This calculation is based on analysis by Signal65 and Futurum Group, assuming a two-year deployment for general knowledge, sales, and software development workloads. Dell notes that while agentic AI is expected to drive 24x token consumption, inference costs are projected to fall more than 90% by 2030.

It provides control over costs and data that public cloud-only approaches cannot match. By owning the infrastructure, enterprises move from a variable consumption model to a fixed-asset model that scales with usage.

Who this is for: Enterprise IT Infrastructure Managers

Enterprise IT Infrastructure Managers can use the Dell AI Factory with NVIDIA to move AI into production while controlling costs. They should compare the two-year ownership costs of this infrastructure against their current public cloud subscription fees.

AI Developers can use the Dell Deskside Agentic AI Solutions for NVIDIA to scale agentic coding workflows. They should download the Dell Deskside Agentic AI Solutions for NVIDIA from Github to begin local setup.

Data Center Architects can use the Dell PowerEdge M9882 or the DSX POD to manage power and cooling. They should monitor the power consumption of the Dell PowerEdge M9882 to ensure it stays within site limits or compare the heat dissipation of liquid cooling in the DSX POD against existing air-cooled infrastructure.

Telecommunications Infrastructure Engineers can use the NVIDIA Aerial RAN Computer (ARC) framework to process AI at the network edge. They should verify that the NVIDIA OpenShell Reference Architecture—a pre-designed plan for deploying hardware and software—meets the data center’s runtime governance standards.

Who this is not for: Consumer laptop shoppers

Consumer laptop shoppers should skip this content as it focuses on enterprise-grade servers and professional workstations. They should look at consumer-focused Dell Inspiron or XPS reviews instead.

Cloud-only software developers can ignore this hardware-centric news as it focuses on on-premise infrastructure ownership. They should instead consult cloud provider documentation for managed AI services.

A scenario for Amara’s software firm

Amara runs a mid-sized software firm that wants to move their coding assistant from a cloud API to local infrastructure to control costs and data. They evaluate the Dell AI Factory with NVIDIA and choose the Pro Precision with GB300 for their dev team because it offers 748GB of coherent memory, allowing them to run frontier open-weight models like Reflection’s Beam locally. The team successfully deploys agentic coding workflows on hardware they own, achieving the 90% cost savings over cloud-only alternatives.

If Amara’s firm required high-throughput inference for millions of concurrent users, the decision flips. They would instead deploy the PowerEdge XE9810 with Groq LPUs to complement their existing GPU clusters for production inference.

Applications for the PowerEdge M9882 and Pro Precision GB300

Cost savings over public cloud and siloed systems

The Dell AI Factory replaces public cloud-only infrastructure, offering up to 90% cost-effectiveness over two years and better data control. Additionally, the NVIDIA Aerial RAN Computer (ARC) framework replaces siloed systems for RAN and AI, allowing those workloads to share the same compute footprint.

Next steps and unconfirmed details

Several components have specific availability dates: the Dell Pro Precision with GB10 (64GB configuration) is available on October 23, while AI observability with Dynatrace and computer vision with Fogsphere will launch in Q4 CY26. The PowerEdge M9882, XE9810, and Dell Managed Services for AI are all scheduled for Q1 CY27. The Dell Modular Data Center with NVIDIA DSX Reference Design will be available in early 2027.

To verify the 90% cost-effectiveness claim

The source did not say who made the specific decisions regarding these product rollouts. Furthermore, the exact availability date for the NVIDIA Aerial RAN Computer (ARC) framework remains unconfirmed. To verify the 90% cost-effectiveness claim, readers can review the ‘Economics of Agentic AI’ analysis by Signal65 and Futurum Group.

Source: Dell Technologies’s press release, as carried by Dell Technologies on 21 October 2026: Dell Technologies Advances Enterprise Agentic AI from the Deskside to Data Center with NVIDIA.