Agentic AI inference at one-tenth the associated fee per token with NVIDIA Vera Rubin NVL72. Agent sandboxes run 50% sooner on NVIDIA Vera than conventional CPUs — whereas enterprise information queries are as much as 3x sooner with the Vera CPU. And 5,000 enterprises like Lilly, Samsung and Honeywell are operating AI workloads on Dell AI Factories with NVIDIA, turning ambition into manufacturing at scale.
That’s the image Michael Dell painted Monday morning at Dell Applied sciences World. Dell sized the stakes: Worldwide AI infrastructure spending may attain $3-4 trillion by 2030, with token consumption projected to develop 3,400% in the identical window.
“There’s a huge AI funding growth thats already underway, and a productiveness growth is starting, and in some corporations, together with ours,” Dell mentioned. “The speed of change has gone parabolic, and it’s not slowing down.”
Then, the Dell chairman and CEO welcomed NVIDIA founder and CEO Jensen Huang to the keynote stage — with a have a look at the NVIDIA portfolio behind him, from a deskside Dell Professional Max with GB10 workstation to a Dell PowerRack with NVIDIA Vera Rubin NVL72.
“We’ve now arrived on the period of helpful AI, which is the rationale why demand goes parabolic, totally parabolic,” Huang mentioned. “What took months now takes weeks. What took weeks now takes days. And what takes days now takes hours. It’s a giant deal in productiveness, however a huge leap in computation necessities.”
The message: Enterprise AI has moved previous pilots into agentic AI and inference deployments at scale. The platform for what’s subsequent is the Dell AI Manufacturing facility with NVIDIA — operating frontier fashions and autonomous brokers securely behind the enterprise perimeter.
A New AI Manufacturing facility for the Agentic Period
The accelerated computing information leads the refresh: The Dell PowerEdge XE9812, constructed on NVIDIA Vera Rubin NVL72, delivers as much as 10x decrease cost-per-token than NVIDIA Blackwell for massive-scale agentic AI inferencing.
It’s joined by PowerEdge XE9880L, XE9885L and XE9882L servers — the primary Dell methods constructed on NVIDIA HGX Rubin NVL8, supporting as much as 144 GPUs per rack with 100% direct liquid-cooled compute nodes and as much as 5.5x the efficiency of HGX B200.
As well as, networking will get the brand new Dell PowerSwitch portfolio with NVIDIA Quantum-X800 InfiniBand, that includes liquid-cooled, co-packaged optics and NVIDIA Spectrum-6 Ethernet.
Dell can also be introducing Dell PowerRack, a totally built-in system – compute, networking and storage engineered as one – with thermal design, energy administration and software program optimization constructed to work collectively from the bottom up. The result’s accelerated AI and high-performance computing workloads at enterprise scale, with out the combination overhead of part meeting.
On the CPU facet, Dell PowerEdge M9822 and R9822 servers convey NVIDIA Vera CPUs to the enterprise AI manufacturing facility. Function-built for agentic AI, Vera runs information pipelines, analytics, sandboxed instruments and code workloads the place every step waits on the final.
With 1.2 TB/s reminiscence bandwidth and predictable efficiency below load, Vera completes agentic workloads 50% sooner than x86 processors, serving to PowerEdge methods enhance AI manufacturing facility output with sooner agent responses and shorter suggestions loops.
“Vera CPU has the best single-threaded efficiency of any CPU on the planet,” Huang mentioned. “It has thrice the reminiscence bandwidth — in consequence, Starburst, DuckDB, all these databases run extremely quick, as a result of the brokers are pounding on the databases, so the CPU had higher be tremendous quick.”
Starburst, a brand new information engine within the Dell AI Information Platform with NVIDIA, delivers 3x sooner question throughput on NVIDIA Vera CPU for large-scale SQL analytics.
Enterprise information offers the gasoline for the AI manufacturing facility. Dell’s replace to its AI Information Platform with NVIDIA facilities on accelerated information engines constructed on NVIDIA CUDA-X libraries — together with cuDF for structured information and cuVS for unstructured information.
A number of clients for Dell AI Manufacturing facility with NVIDIA have been featured within the keynote.
Diogo Rau, govt vp and chief info and digital officer of Lilly, joined early within the keynote — discussing Lilly’s AI-driven developments and innovation in life sciences, powered by AI infrastructure deployed at scale with Dell and NVIDIA.
He described know-how as key to delivering cutting-edge science, at scale. “I believe we’re on the verge of possibly having the ability to finish illness as we all know it,” Rau mentioned. “One thing like that was fully unimaginable 20 years in the past, however immediately we are able to think about it.”
A video from Samsung adopted — highlighting use circumstances for R&D chip design and manufacturing operating on Dell AI Manufacturing facility with NVIDIA.
Honeywell chief know-how officer Suresh Venkatarayalu joined Michael Dell to stroll by means of the corporate’s transfer from public cloud to on-premises AI — utilizing the Dell AI Manufacturing facility and Dell AI Information Platform with NVIDIA for industrial AI use circumstances, digital twins and automation from information heart to the sting.
“For me, partnering with Dell and NVIDIA isn’t just about getting infrastructure,” Venkatarayalu mentioned. It’s the complete AI stack, he defined: scalable, secured and trusted by clients.
And in monetary companies, Hudson River Buying and selling, the algorithmic buying and selling agency, is increasing its Dell deployment to energy AI-driven analysis — operating Dell PowerEdge XE9685L servers with NVIDIA accelerated computing and NVIDIA Spectrum-X Ethernet to scale with the agency’s information, fashions and ambition.
Brokers and Fashions on Premises — Securely
Dell’s personal AI adoption survey, cited from the keynote stage, discovered that 67% of AI workloads now run exterior the cloud — on premises, on gadget, on the edge or in colocation — and that 88% of respondents are operating not less than one AI workload on premises.
The on-premises AI bulletins answered a query Dell put to the room instantly: “How do you deploy the world’s finest AI fashions the place you want them, with safety and governance inbuilt?”
The reply rests on NVIDIA Confidential Computing — delivered along with Fortanix, Google, Crimson Hat and different companions — as the inspiration for securely deploying frontier fashions contained in the enterprise with out exposing mannequin IP or information.
This allows enterprises to guard AI fashions and delicate information in use whereas capturing the token effectivity, efficiency and value benefits of on-premises AI infrastructure.
Frontier Proprietary Fashions, Protected by Confidential Computing
Google Distributed Cloud (GDC) with Gemini 3.0 is now obtainable in preview on Dell PowerEdge XE9780 servers, accelerated by NVIDIA Blackwell and secured by NVIDIA Confidential Computing — giving enterprises a personal confidential computing setting for superior AI.
SpaceXAI will even convey the most recent SpaceXAI fashions on premises to the Dell AI Manufacturing facility, with NVIDIA Confidential Computing holding mannequin weights and enterprise information protected finish to finish.
Open Frontier Fashions, Working Natively on the Dell AI Manufacturing facility
NVIDIA Nemotron fashions — frontier-class open intelligence — run on Dell AI Manufacturing facility infrastructure for enterprises that need open weight fashions tuned to their very own domains and information.
Reflection’s open supply frontier AI fashions are additionally approaching premises, purpose-built for regulated industries, governments and sovereign entities.
Further open fashions — MiniMax-M2.7, DeepSeek Professional, DeepSeek-V4, GLM 5.1 and Kimi K2.6 with NVIDIA NVFP4 optimization — can be found on the Dell Enterprise Hub on Hugging Face, becoming a member of Gemma 4, NVIDIA Nemotron Tremendous 3, Mistral Small 4 and Arcee Trinity-Massive-Considering.
On this new agentic period, enterprises additionally want brokers to work securely throughout the hybrid and on-premises environments the place their information, methods and workflows already dwell.
OpenAI Codex will join with the Dell AI Information Platform, to assist clients convey Codex nearer to the interior context that makes brokers helpful: codebases, documentation, enterprise methods, operational information and group workflows. Dell and OpenAI will even discover how Codex can join with the Dell AI Manufacturing facility.
Ecosystem of Software program Companions
Dell introduced a number of new software program partnerships for frequent enterprise AI use circumstances starting from agentic AI and code assistants to pc imaginative and prescient.
This contains Palantir’s sovereign AI OS reference structure with NVIDIA, introduced in March, which now runs on Dell infrastructure — for on-premises deployment of Palantir Ontology and AIP, built-in with the NVIDIA Sovereign AI OS Reference Structure.
As well as, ServiceNow clients will be capable of leverage the Dell AI Manufacturing facility to convey collectively infrastructure and enterprise workflow automation, enabling organizations to find, govern and operationalize AI targeted on enterprise outcomes.
Dell additionally introduced new options with a variety of AI leaders and software program innovators —together with Fogsphere, Ipsotek, Mistral AI, Poolside and Uneeq — in addition to safety partnerships with CrowdStrike and Fortanix.
Brokers From Deskside to Information Middle
Essentially the most private information: Dell Deskside Agentic AI with the NVIDIA NemoClaw stack, NVIDIA OpenShell runtime and NVIDIA Nemotron open fashions run on Dell Professional Max with GB10 and GB300 powered by the NVIDIA Grace Blackwell structure, in addition to Dell Professional Precision methods powered by NVIDIA RTX PRO Blackwell workstation platforms.
The customization layer: NVIDIA Nemotron, NVIDIA Agent Toolkit and NVIDIA NeMoClaw — the agent orchestration harness Huang described on stage because the connective layer between native fashions and enterprise information — present the inspiration for constructing enterprise autonomous brokers, enabling organizations to customise fashions, orchestrate agent workflows and securely join brokers to enterprise information and instruments.
The safety layer: NVIDIA OpenShell — an open supply runtime for improvement and deployment of autonomous brokers with safety and privateness controls — allows company coverage enforcement on the infrastructure layer and is built-in with main enterprise software program platforms.
NVIDIA OpenShell is now supported throughout the whole Dell AI Manufacturing facility with NVIDIA, giving builders a safe runtime to construct, deploy and govern AI brokers from workstations to servers.
Dell additionally highlighted assist for the NVIDIA AI-Q Blueprint, giving enterprises a reference instance to deploy multi-agent workflows for deep analysis — accelerating the trail from improvement to pilot to manufacturing.
Day Two
On day two of Dell Applied sciences World, Dell’s Chief Operations Officer Jeff Clarke and Infrastructure Options Group President Arthur Lewis will go deeper on NVIDIA Vera CPU, Vera Rubin, Confidential Computing and Nemotron — with a dwell demo of Dell Deskside Agentic AI.
The themes Huang and Dell set Monday — secure, long-running brokers, full-stack factories, safe on-premises deployment — arrange the broader bulletins NVIDIA will convey to GTC Taipei at COMPUTEX, operating June 1-4.
