Dell 与 NVIDIA 扩展 AI 工厂：采用企业超6,500家、PowerEdge M9882 单机架装256颗 Vera CPU
Dell Technologies 10月21日在 Berlin 举行的 NVIDIA GTC 大会上宣布扩展 Dell AI Factory with NVIDIA 产品线，涵盖 AI 解决方案、桌面端智能体方案与专用 AI 基础设施，已有超过6,500家企业采用该平台，较上季度的5,000家增加，新增服务器与模块化数据中心多数定于2027年初上市。
Dell 称，企业级智能体 AI 预计到2028年将成为单一最大工作负载，到2030年增长速度接近生成式 AI 的三倍；同期推理成本预计下降超过90%，而智能体 AI 将带来24倍的 token 消耗。公司此次把智能体的执行与模型推理在硬件上分开：新款 PowerEdge M9882 服务器在单个48U机架内装入256颗 NVIDIA Vera CPU，专用于智能体的编排、工具调用、代码执行与评估；PowerEdge R8822 把双路 Vera CPU 的风冷性能带入标准19英寸机架，支持最多四块双宽 GPU，本月上市；PowerEdge XE9810 则配备16个直液冷计算托盘、采用 Groq 3 语言处理单元做机架级推理，定于2027年第一季度上市。
桌面端方面，Dell Deskside Agentic AI 新增经验证的部署模板，写明硬件、模型与软件栈，可从单个开发者扩展到1,000台以上系统的企业级部署；Dell Pro Precision with GB10 的64GB统一内存版本10月23日上市，Pro Precision with GB300 的 Windows 版本将提供最高748GB一致性内存与20 petaflops 的 FP4 算力。基础设施方面，采用 NVIDIA DSX POD 的 Dell 模块化数据中心为集装箱式参考设计，集成计算、存储、网络、机架级供电与液冷，定于2027年初上市；Dell AI 托管服务定于2027年第一季度推出。
Dell Technologies 董事长兼首席执行官 Michael Dell 说，AI 基础设施已不再是成本中心，而是价值创造中心，企业正扩大并重新分配预算，因为 AI 基础设施带来增长、生产力与竞争优势。NVIDIA 创始人兼首席执行官 Jensen Huang 说，AI 智能体会推理、使用工具并学习，它们需要一种新型计算机；两家公司正从桌面端到数据中心共同打造 AI 时代的计算基础，让企业以工业规模安全地构建与运行 AI 智能体。
Dell and NVIDIA Expand AI Factory: Over 6,500 Enterprise Customers, PowerEdge M9882 Packs 256 Vera CPUs in a Single Rack
Dell Technologies announced on October 21 at NVIDIA GTC in Berlin a broad set of advancements across the Dell AI Factory with NVIDIA, spanning AI solutions, deskside agentic solutions and purpose-built AI infrastructure, with more than 6,500 enterprise customers now on the platform, up from 5,000 last quarter, and most of the new servers and modular data centers scheduled for early 2027.
Dell said enterprise agentic AI is expected to be the single largest workload by 2028, growing nearly three times faster than generative AI by 2030, while inference costs are projected to fall more than 90 percent by 2030 and agentic AI is expected to drive 24 times the token consumption. The new hardware separates agent execution from model inference: the PowerEdge M9882 packs 256 NVIDIA Vera CPUs into a single 48U rack, purpose-built for agent orchestration, tool use, code execution and evaluation; the PowerEdge R8822 brings air-cooled two-socket Vera CPU performance to standard 19-inch racks with support for up to four double-wide GPUs and is available this month; and the PowerEdge XE9810, featuring 16 direct-liquid-cooled compute trays powered by Groq 3 Language Processing Units for rack-scale inference, will be available in the first quarter of 2027.
On the deskside, Dell Deskside Agentic AI adds validated reference templates specifying hardware, models and software stacks, scaling from a single developer to enterprise deployments of 1,000 or more systems. The Dell Pro Precision with GB10 in a 64GB unified memory configuration became available on October 23, and the Pro Precision with GB300 on Windows will offer up to 748GB of coherent memory and 20 petaflops of FP4 compute. The Dell Modular Data Center with NVIDIA DSX POD is a containerized reference design integrating compute, storage, networking, rack-level power and liquid cooling, available in early 2027, and Dell Managed Services for AI will be available in the first quarter of 2027.
"Enterprises used to think about AI infrastructure as an IT budget question, but that's changed. AI infrastructure is no longer a cost center. It's a value creation center," said Michael Dell, chairman and chief executive officer of Dell Technologies. Jensen Huang, founder and chief executive officer of NVIDIA, said AI agents reason, use tools and learn, and need a new kind of computer, adding that the two companies are building the computing foundation for the age of AI, from the deskside to the data center, so companies can build and run AI agents securely at industrial scale.
来源：GSM·美国（原文链接）
图片来源：Dell Technologies 总部大楼。照片：Dell Technologies，CC BY-SA 4.0，经 Wikimedia Commons
