Dell and NVIDIA Revolutionise In-House AI with AI Factory

Dell Technologies has announced advancements across the Dell AI Factory with NVIDIA for businesses that are increasingly choosing to develop AI capabilities in-house and on-premises. 

The Dell AI Factory with NVIDIA was launched in March 2024 and has since had more than 4,000 customer deployments with early adopters seeing up to 2.6x ROI within the first year. 

It is powered by NVIDIA GPUs, networking and AI software to handle retrieval-augmented generation and multimodal search for agentic workflows and large-scale data processing.

The updates can be categorised into helping four categories: desktop AI development and autonomous agents, production AI at scale, enterprise workloads in the data centre, and high-performance networking and emerging technologies. 

The development of desktop AI and autonomous agents

The Dell Pro Max with GB10 and GB300 are essentially supercomputers that sit on your desk. They are purpose-built to handle complex, long-running AI programmes and can learn and adapt as they work.

They provide the massive computing power, memory and reliability needed to develop and test AI at a “trillion-parameter scale”, according to Dell. 

Dell is the first manufacturer to incorporate the extremely powerful NVIDIA GB300 Grace Blackwell chip into a desktop machine.

“We’re the first OEM to ship the GB300 Grace Blackwell Ultra Desktop Superchip with our Dell Pro Max desktop, bringing enterprise-grade AI computing directly to developers’ desks,” says Arthur Lewis, President of Infrastructure Solutions Group for Dell Technologies.

Plus, the new Dell Pro Precision workstations are designed as next-generation powerhouse computers for AI developers and data scientists.

They come in different formats to suit various needs, including a tower configuration that can hold up to five NVIDIA RTX PRO Blackwell Desktop Generation GPUs and laptops that still deliver significant performance.

Enhancing production of AI at scale

Dell’s PowerEdge servers are helping at the data centre level.

They deliver the performance needed for training large language models and running real-time inference for autonomous AI agents. 

PowerEdge XE9812 uses the NVIDIA Vera Rubin NVL72 platform to handle tasks like real-time training and inference, such as getting instant results from trained AI models. 

PowerEdge XE9880L, XE9882L and XE9885L are a series of water-cooled servers that use the NVIDIA HGX Rubin NVL8.

Their design focuses on improving AI performance within a company’s existing server rooms and without requiring more electricity.

Improving enterprise workloads in the data centre

The Dell PowerEdge R770, R7715 and R7725 servers allow companies to easily inject AI capabilities into their existing, everyday server setups.

By configuring them with the new NVIDIA RTX PRO 4500 Blackwell Server Edition GPU, teams can add specialised AI speed and processing power to their standard equipment.

High-performance networking and emerging technologies

Dell has also revealed new components focused on the networking and physical infrastructure required to connect all AI servers within a data centre. 

The Dell PowerSwitch SN6000-series are high-speed network switches that physically connect the AI servers.

They are designed to move huge amounts of data – up to 1.6Tbs – quickly and can be water-cooled, allowing for efficient operation.

Meanwhile, the PowerSwitch SN5610 and SN2201 give data centre teams more flexibility by offering a wider variety of operating systems, including Cumulus Linux and Enterprise SONiC Distribution by Dell Technologies, for managing the network.

NVIDIA Quantum-X800 InfiniBand Q3300-LD is another type of liquid-cooled switch that uses InfiniBand technology to deliver extremely fast, dedicated networking for AI workloads and cloud applications that need very high bandwidth.

Finally, the Dell Integrated Rack Scalable Systems is a physical rack setup where all the servers and network gear are housed. 

“For enterprises still stuck between pilot and production, the lesson is simple: integration matters, data readiness matters, deployment expertise matters,” Lewis says.

“A partner who delivers all three is the difference between AI as an experiment and AI as a business driver.”

These announcements were made at NVIDIA GTC 2026, taking place between 16 and 19 March in San Jose, California.