AI 模型和平台
Vertiv 发布 SmartIT 工程前AI 基础设施平台
在 Google 上的首选来源中添加 Unite.AI
Vertiv Holdings Co 推出 2026年10月26日的Vertiv SmartIT解决方案，是一个基于NVIDIA MGX和Open Compute Project ORV3架构进行高密度AI部署的预设计，工厂验证的基础设施平台。 该平台现在可用于全球的人工智能部署，是为企业、研究机构和现有设施和新数据中心的合用同一地点提供者设计的。
可扩展平台支持在不重新设计的情况下，在现有的企业设施、新建筑、商业空间、同地办公和超规模环境中部署。 根据公告，SmartIT此前与NVIDIA一同部署，以支持NVIDIA DGX GB300,Vertiv称其为企业可用的最高性能，规模最大的AI系统。 该设计将机架基础设施、DC电力转换和配电、直对芯片液冷却、机架冷却分配和监测结合起来，这些系统在装运前在工厂验证，并在现场与设施动力和冷却基础设施连接。
" AI正在从根本上改变对数据中心基础设施的要求，要求更密切地协调各种电力、冷却和信息技术系统，以支持日益复杂的工作量。 " He 称 SmartIT 带来这些学科在一个工厂验证的架构中结合在一起，结合了Vertiv的关键基础设施专业知识来帮助客户简化部署，在他们的AI环境中建立更大的一致性，并建立一个可以与他们的计算需求一起进化的基础。
支持的平台和Rack组成
根据公告，SmartIT支持NVIDIA GB300 NVL72,NVIDIA Vera Rubin NVL72,NVIDIA DGX Rubin NVL8和NVIDIA HGX Rubin NVL8系统在NVIDIA MGX和OCP ORV3架构上。 Vertiv 称平台在整合动力，冷却，监测和控制的同时，在同一个共同架构内从单一架子到多架部署的尺度上进行内置监测，漏泄检测和支持Redfish,SNMP和Modbus协议。
Vertiv的SmartIT产品页面指出，该平台将Vertiv PowerDirect,DC巴士巴的发售量高达1400安培，机架内多路和Vertiv CoolChip CDU组合成一个单一的预先设计并经过验证的软件包。 冷却选项包括一个机柜内机床，一个机柜内冷却剂分配装置(CDU)和一个后门热交换器。 综合监测横跨动力架，CoolChip CDU和Vertiv CoolLoop RDHx(雷亚尔-门热交换器)，具有主动的漏气探测和自动散射防护。
产品页列出NVIDIA GB300 NVL72在NVIDIA MGX架构上的SmartIT,OCP ORv3在OCP ORV3 HPR V1架构上的SmartIT. 第三个型号是SmartIT for NVIDIA Vera Rubin，在MGX的Vera Rubin NVL72和Rubin NVL8平台上被列出为即将到来。 列出的单位高90.23英寸，包括铸造机，宽23.6英寸，深42.05英寸，或47.25英寸，有延伸，重416.7磅。 SmartIT持有一年的有限保修权，但必须遵守Vertiv的标准条款、条件和排除条件，Vertiv列出了涵盖银行、金融服务和保险、数据中心、同地办公和托管、教育、政府、医疗保健以及零售和批发的纵向保修。
订购模式、服务和GTC柏林示范
Vertiv 称每个SmartIT配置在系统层面进行预先设计并验证，作为一个解决方案订购并作为一个单元交付，从一个单机架部署到整个企业，合用地点和超规模站点的全吊舱。 模型称公司以工厂验证取代现场集成，将委托范围限制在设施连接上，从而压缩了部署时间表，并给客户一个单次订单，一个合作伙伴和一个建筑跨越架，动力，冷却和监测。 Vertiv还拥有称棕田企业设施和A级商业建筑，可以支持AI级密度而无需完全重建，同样的建筑块尺度从单一架子到多架子的部署在各阶段之间没有重新设计。
Vertiv全球服务支持该平台的部署、调试和生命周期管理。 Vertiv 称 SmartIT也可以与互补产品一起部署：Vertiv SmartRun将集成延伸至全行，Vertiv的机架动力分配单元提供电源分配和监测，Vertiv Unify提供跨电力和热力基础设施的综合监测，分析和控制，包括泄漏检测和协调反应控制。
Vertiv于2026年10月在NVIDIA GTC柏林展示一个SmartIT单元与NVIDIA HGX Rubin NVL8演示单元对齐。 Vertiv 称的演示呈现了一个综合基础设施架构，旨在支持NVIDIA DGX Rubin NVL8的部署，使得客户能够以合适的规模开始，并随着AI需求的增长在共同的基础设施架构内扩展。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
AI Models & Platforms
Vertiv Launches SmartIT Pre-Engineered AI Infrastructure Platform
Add Unite.AI to your preferred sources on Google
Vertiv Holdings Co introduced the Vertiv SmartIT solution on October 26, 2026, a pre-engineered, factory-validated infrastructure platform for high-density AI deployments based on NVIDIA MGX and Open Compute Project ORV3 architectures. The platform is available now for AI deployments globally and is designed for enterprises, research institutions and colocation providers across existing facilities and new data centers.
Vertiv said the scalable platform supports deployment across existing enterprise facilities, new builds, commercial spaces, colocation and hyperscale environments without redesign. According to the announcement, SmartIT was previously deployed with NVIDIA to support the NVIDIA DGX GB300, which Vertiv described as the highest-performing, largest-scale AI system available for enterprises. The design integrates rack infrastructure, DC power conversion and distribution, direct-to-chip liquid cooling, in-rack cooling distribution and monitoring, with those systems validated at the factory before shipment and connected to facility power and cooling infrastructure on site.
“AI is fundamentally changing the demands placed on data center infrastructure, requiring closer coordination across power, cooling and IT systems to support increasingly complex workloads,” said Scott Armul, Vertiv’s Chief Product and Technology Officer. He said SmartIT brings those disciplines together in one factory-validated architecture, combining Vertiv’s critical infrastructure expertise to help customers simplify deployment, build greater consistency into their AI environments and establish a foundation that can evolve alongside their computing needs.
Supported Platforms and Rack Composition
According to the announcement, SmartIT supports NVIDIA GB300 NVL72, NVIDIA Vera Rubin NVL72, NVIDIA DGX Rubin NVL8 and NVIDIA HGX Rubin NVL8 systems on NVIDIA MGX and OCP ORV3 architectures. Vertiv said the platform scales from a single rack to multi-rack deployments within one common architecture while integrating power, cooling, monitoring and controls, with built-in monitoring, leak detection and support for the Redfish, SNMP and Modbus protocols.
Vertiv’s SmartIT product page states that the platform combines Vertiv PowerDirect, DC busbar distribution up to 1400 amps, in-rack manifolds and the Vertiv CoolChip CDU into a single pre-engineered and validated package. Cooling options include an in-rack manifold, an in-rack coolant distribution unit (CDU) and a rear-door heat exchanger. Integrated monitoring spans the power shelves, the CoolChip CDU and the Vertiv CoolLoop RDHx (rear-door heat exchanger), with proactive leak detection and automated shunt-trip protection.
The product page lists SmartIT for NVIDIA GB300 NVL72 on the NVIDIA MGX architecture and SmartIT for OCP ORv3 on the OCP ORV3 HPR V1 architecture. A third model, SmartIT for NVIDIA Vera Rubin, is listed as coming soon for the Vera Rubin NVL72 and Rubin NVL8 platforms on MGX. Listed units stand 90.23 inches high including castors, 23.6 inches wide and 42.05 inches deep, or 47.25 inches with an extension, and weigh 416.7 pounds. SmartIT carries a one-year limited warranty subject to Vertiv’s standard terms, conditions and exclusions, and Vertiv lists target verticals spanning banking, financial services and insurance; data center, colocation and hosting; education; government; healthcare; and retail and wholesale.
Ordering Model, Services and the GTC Berlin Demonstration
Vertiv said each SmartIT configuration is pre-engineered and validated at the system level, ordered as one solution and delivered as one unit, deployable from a single rack to a full pod across enterprise, colocation and hyperscale sites. The company said the model compresses deployment timelines by replacing on-site integration with factory validation, limiting commissioning scope to facility connections, and gives customers a single order, one partner and one architecture spanning rack, power, cooling and monitoring. Vertiv also said brownfield enterprise facilities and Class A commercial buildings can support AI-class density without full reconstruction, and that the same building block scales from a single rack to multi-rack deployments without redesign between phases.
Vertiv Global Services supports deployment, commissioning and lifecycle management for the platform. Vertiv said SmartIT can also be deployed alongside complementary products: Vertiv SmartRun extends integration across the row, Vertiv rack power distribution units provide power distribution and monitoring, and Vertiv Unify delivers integrated monitoring, analytics and controls across power and thermal infrastructure, including leak detection and coordinated response control.
Vertiv showcased a SmartIT unit paired with an NVIDIA HGX Rubin NVL8 demonstration unit at NVIDIA GTC Berlin in October 2026. Vertiv said the demonstration presented an integrated infrastructure architecture designed to support NVIDIA DGX Rubin NVL8 deployments, allowing customers to start at the right scale and expand within a common infrastructure architecture as AI demand grows.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.