At Dell Technologies World, Jensen Huang Says "Demand Is Growing Explosively"

Michael Dell and Jensen Huang emphasized on the Dell Technologies World stage that AI demand and infrastructure investment are already growing on a parabolic curve. Dell projected that global AI infrastructure spending could reach $3 trillion to $4 trillion by 2030, and that token consumption could grow by 3,400% over the same period. Already, 5,000 companies, including Lilly, Samsung, and Honeywell, are running AI workloads on the Dell AI Factory with NVIDIA.

The core announcement was the expansion of the Dell AI Factory with NVIDIA.

• The PowerEdge XE9812, based on NVIDIA Vera Rubin NVL72, was touted to lower the per-token cost of agentic AI inference by up to 10x.
• The XE9880L, XE9885L, and XE9882L are Dell's first systems based on HGX Rubin NVL8, targeting up to 144 GPUs per rack, full direct liquid cooling, and performance of up to 5.5x compared to HGX B200.
• The PowerSwitch portfolio includes Quantum-X800 InfiniBand and Spectrum-6 Ethernet, and PowerRack was presented as a fully integrated rack combining compute, network, and storage.

On the CPU side, Vera CPU powers agentic workloads. The PowerEdge M9822/R9822, built on 1.2 TB/s memory bandwidth and predictable performance, was presented as delivering 50% faster agent execution compared to x86, and Huang said Vera CPU has world-class single-threaded performance. Starburst stated that large-scale SQL query throughput is 3x faster on Vera CPU. The Dell AI Data Platform is also expanding with a CUDA-X-based data engine, leveraging cuDF and cuVS.

According to a Dell survey, 67% of AI workloads are already running outside the cloud, and 88% of respondents are operating at least one AI workload on-premises. In line with this trend, Dell put NVIDIA Confidential Computing front and center, emphasizing an approach—together with Fortanix, Google, and Red Hat—of deploying frontier models while protecting model IP and data.

The model and partner ecosystem also expanded. Google Distributed Cloud with Gemini 3.0 is available as a preview on the XE9780, and SpaceXAI models are also coming on-premises. In addition to open models from NVIDIA Nemotron and Reflection, MiniMax-M2.7, DeepSeek Pro, DeepSeek-V4, GLM 5.1, and Kimi K2.6 have been added to the Dell Enterprise Hub on Hugging Face. The partner ecosystem now includes Palantir, ServiceNow, Mistral AI, Poolside, Uneeq, CrowdStrike, Fortanix, and others.

On the desktop side, the combination of NemoClaw, OpenShell, and Nemotron is coming to Dell Pro Max GB10/GB300 and RTX PRO Blackwell workstations, supporting agent development, deployment, and governance. Dell also announced support for AI-Q Blueprint, providing a reference implementation for multi-agent workflows, and said it would connect OpenAI Codex with the Dell AI Data Platform to build agents that are more closely tied to internal codebases and business systems. This momentum continues into announcements at GTC Taipei at COMPUTEX, held June 1–4.