Live from Dell World 26: Key Takeaways from Day 1 + Keynote

研报英文原文证据摘录

Live from Dell World 26: Key Takeaways from Day 1 + Keynote

orchestration stack, OpenShell runtime and Nemotron open-weight

models that’ll let developers test, build and fine-tune agents locally on

70-250B parameter models (and up to 1T parameters), with DELL

claiming break-even versus public cloud APIs in as little as 3 months.

Model Announcements: DELL has expanded model choice within its

AI Factory, starting with a refreshed Dell Enterprise hub on Hugging

Face that’s adding support for MiniMax-M2.7, DeepSeek Pro,

DeepSeek V4, GLM-5.1 and Kimi K2.6 (with NVIDIA NVFP4

optimization) to an existing roster that includes NVIDIA Nemotron

Super 3, Mistral Small 4, Gemma 4 and Arcee Trinity-Large-Thinking.

In addition, DELL announced a series of frontier-model partnerships

that’ll bring closed and proprietary models on-prem to the Dell AI

Factory for confidential computing environments. The announcements

include: 1) Google Distributed Cloud with Gemini 3.0 running on

PowerEdge XE9780; 2) OpenAI's Codex agentic harness and GPT-5.5

connectivity through the Dell AI Data Platform; 3) xAI's Grok available

on-prem or hybrid as a secure enterprise-grade assistant; and 4)

Palantir's Foundry AI Platform and Ontology deployed on ObjectScale

and PowerFlex.

NVIDIA-related Announcements: Dell announced the PowerEdge

XE9812 built on NVIDIA's Vera Rubin NVL72, claiming up to 10x lower

cost per token versus Blackwell for agentic AI inferencing. Dell also

introduced PowerEdge XE9880L, XE9885L and XE9882L servers, the

first Dell systems built on NVIDIA HGX Rubin NVL8, supporting up to

144 GPUs per rack with 100% direct liquid-cooled compute nodes and

up to 5.5x the performance of the prior HGX B200 generation. On the

CPU side, new PowerEdge M9822 and R9822 servers bring NVIDIA's