Volume 343, Issue 3 · 10 articles
At NVIDIA GTC, Dell announces a Modular Data Center and new PowerEdge M9882, R8822 and XE9810 AI platforms.
Dell used NVIDIA GTC to announce four additions to its AI infrastructure portfolio.
The Dell Modular Data Center, built on the NVIDIA DSX POD reference design, is a factory-integrated container that Dell says can cut deployment at a prepared site from six months to 14 days.
The PowerEdge M9882 packs 256 NVIDIA Vera CPUs into one rack for agentic AI orchestration, and the R8822 is an air-cooled 2U server for standard racks.
The PowerEdge XE9810 is Dell's first rack-scale inference platform built on NVIDIA Groq LPUs.
Dell outlines its cyber resilience approach of securing infrastructure, detecting compromise and recovering clean data quickly.
Dell describes a cyber resilience approach built on securing infrastructure, detecting attacks and recovering operations without reintroducing malicious code.
Protection relies on defense-in-depth measures such as immutable, encrypted backups, tight access control, a secure supply chain and an isolated vault.
Detection layers include Dell Managed Detection and Response delivered with CrowdStrike and anomaly detection in PowerProtect Data Manager.
Dell Cyber Detect inspects PowerStore and PowerMax snapshots and vaulted backups to find clean recovery copies, with the best path depending on how far a breach has spread.
Dell launches Deskside Agentic AI Solutions for NVIDIA, five validated blueprints that scale from one workstation to 1,000+ systems.
Dell has launched Dell Deskside Agentic AI Solutions for NVIDIA, five validated blueprints that bundle recommended hardware, software and models.
The blueprints cover Agentic Coding, Agentic VSS, NemoClaw, Hermes on NemoClaw and a Higher Education AI Lab.
Each can scale from a single developer workstation to more than 1,000 systems.
Dell also expanded its lineup with a 64GB Dell Pro Precision with GB10 and made the Dell Pro Precision with GB300 available with Windows.
Dell adds SUSE's Linux AI Factory stack as an operating system option for Dell AI Factory with NVIDIA, expected in Q1 2027.
Dell is adding SUSE's Linux AI Factory software stack as an operating system option for Dell AI Factory with NVIDIA deployments.
Dell says many AI pilots stall because models, tools, security requirements and infrastructure keep changing, leading to fragmented setups.
The SUSE stack provides a hardened OS foundation on a validated reference architecture for private generative AI, RAG and agentic workflows, including sovereign and air-gapped environments.
Dell estimates availability in Q1 2027.
Dell explains how its PowerEdge servers turn NVIDIA's Aerial RAN Computer reference design into deployable telecom infrastructure.
Dell says its PowerEdge servers align with NVIDIA's Aerial RAN Computer, an open reference design combining CPUs, NVIDIA GPUs and NICs.
ARC comes in a 2U version for mobile switching offices and aggregated sites and a denser 4U version for larger centralized RAN deployments.
Dell argues running RAN and AI workloads on shared infrastructure lowers TCO, reduces silos and supports zero-touch automation.
The post describes work with NVIDIA and Nokia, with Nokia providing RAN software and radio technology.
Dell proposes that state and local governments place modular data centers at utility sites with unused power capacity.
Dell suggests state and local governments could deploy modular data centers at existing utility sites with stranded power capacity.
Pre-built, pre-tested units can be delivered quickly and may avoid new generation or major grid upgrades.
Because many government AI tasks need real-time results, the infrastructure can be smaller and located close to the people it serves.
Dell says closed-loop cooling can reduce water use and cooling energy, and its modular units include power monitoring and demand-response features.
Dell and Mistral announce a validated solution for running Mistral Studio on the Dell AI Factory with NVIDIA.
Dell and Mistral announced a validated solution for running Mistral Studio on the Dell AI Factory with NVIDIA, timed to GTC Berlin.
Dell argues moving AI to production requires a foundation connecting infrastructure, data, applications and operations.
Mistral Studio provides one environment to build, customize, deploy and govern AI applications with RAG, agent workflows, tool calling and observability.
The approach emphasizes deployment across on-premises, edge and private cloud, consistent governance and customer ownership of models and knowledge.
Dell and Reflection unveil a validated solution to run Reflection's open-weight Beam model on Dell AI Factory with NVIDIA.
Dell and Reflection unveiled a validated solution at NVIDIA GTC Berlin 2026 that deploys Reflection's open-weight Beam model inside customers' own environments.
Beam is a sparse Mixture-of-Experts model with 501 billion total parameters, 23 billion active per token and a context window of up to one million tokens.
The stack runs on PowerEdge servers with NVIDIA GPUs, PowerSwitch networking, PowerScale storage and Dell Services, and supports fine-tuning on proprietary data.
Beam's weights are released under an Apache 2.0 license.
Maya HTT built M-Bot, an industrial AI assistant on Dell AI Factory with NVIDIA, to search decades of engineering data securely.
Dell profiles M-Bot, an industrial AI assistant built by Canadian engineering firm Maya HTT, which has more than 40 years of experience.
Maya HTT built M-Bot on Dell AI Factory with NVIDIA so engineering and manufacturing teams can search decades of technical data without moving sensitive information off premises.
M-Bot answers questions with links to supporting sources so users can verify information.
Maya HTT reports significant time savings and faster onboarding and plans to extend M-Bot toward agentic workflows.
Dell's Shamus Jones says edge AI pays off when workloads sit at the right layer, with security, reliability and lifecycle cost in mind.
Computing is moving from centralized cloud toward the edge, but Dell's Shamus Jones says bandwidth savings and legacy constraints, more than latency, usually drive edge projects, and remote manageability soon matters as much as speed.
Far-edge devices offer immediate action but face theft, tampering and power limits; the near-edge offers the best balance; the far cloud offers scale at higher latency.
Placing workloads at the wrong layer is a costly mistake.
Jones cites STL Partners' forecast of a $445 billion edge market by 2030 and says small quantized models, such as a tuned YOLO, can run on CPU-only edge servers.