AI-optimized infrastructure for every industry
From hyperscale to healthcare, Vertiv™ 360AI delivers the power, cooling, and expertise to enable the next generation of AI.
Why Vertiv™ 360AI?
Cloud, Hyperscale and Colocation
Close alignment with AI chip providers’ roadmaps, with the ability to customize at scale. Backed by decades of design, build, and commissioning expertise, Vertiv enables lower costs, improved manageability, and faster deployment.
Enterprise
Deep vertical expertise with consulting and support throughout the AI adoption journey. Rapid configuration capabilities and trusted partner ecosystems help enterprises expand quickly from core to edge.
Communications networks
Flexible architectures and turnkey deployment models designed for small cell, macro site, and central office environments. Intelligent systems simplify network management while enabling innovation.
Commercial & industrial
High-level protection for mission-critical operations in healthcare, manufacturing, and transportation. Solutions are tailored to performance and budget needs, supported by expert technical teams ready to respond quickly.
New AI architecture design
Download Vertiv’s energy-efficient cooling and power reference architecture for the NVIDIA GB300 NVL72 platform, available as SimReady assets in NVIDIA Omniverse blueprint.
Key specs:
- Capacity: 2.5 MW | Redundancy: 4 to make 3 (power), N+1 (cooling)
- Rack density: 16 racks @ 142kW, 24 racks @ 10kW
- Cooling: 77% Liquid (Direct-to-Chip), 23% Air
- Power distribution: 33kVA DC shelves, 400A busbars, 500kVA PDUs
- UPS: 1000kVA for IT, 240kVA for facility
- Supports: NVIDIA Blackwell DGX SuperPOD architecture
改造设计
无论是测试新的 AI 模型还是升级本地基础设施，这些可直接订购的改造设计都能简化您的下一次部署。
<
适用于改造的可报价设计
| 总货架密度 | 货架总数 | 总部署规模 | 设计标识 | 冷却技术 | ||
|---|---|---|---|---|---|---|
| 北美洲 | 欧洲、中东及非洲地区 | 亚洲 | ||||
| 40千瓦 | 4 | 160千瓦 | 4X160R |
不适用 | 4X160RA | 气体 |
| 70千瓦 | 1 | 70千瓦 | 1L88R |
1L88RE |
1L88RA |
液体 + 空气 |
| 80千瓦 | 1 | 80千瓦 | 1X80 |
1X80E | 1X80A | 气体 |
| 100kW | 1 | 100kW | 1L100R |
1L100RE |
1L100RA |
液体 + 空气 |
| 100kW | 4 | 400千瓦 | 4L400R |
不适用 | 4L400RA | 液体 + 空气 |
| 100kW | 4 | 400千瓦 | 4XL400 |
4XL400E |
4XL400A |
液体 + 空气 |
| 100kW | 5 | 500千瓦 | 5L500 |
5L500E |
5L500A |
液体 + 空气 |
| 100kW | 12 | 1200千瓦 | 12XL1200 |
12XL1200E |
12XL1200A |
液体 + 空气 |
| 100kW | 14 | 1400千瓦 | 14L1400 |
14L1400E |
14L1400A |
液体 + 空气 |
| 115千瓦 | 1 | 132千瓦 | 不适用 | 1XL115E |
不适用 | 液体 + 空气 |
参考设计
Vertiv 的参考设计为 AI 数据中心构建提供了一个清晰的起点，并简化了高密度工作负载的集成。
新建建筑设计
| 总货架密度 | 货架总数 | 总部署规模 | 参考 GPU 类型 | 设计标识 | 冷却技术 | ||
|---|---|---|---|---|---|---|---|
| 北美洲 | 欧洲、中东及非洲地区 | 亚洲 | |||||
| 20kW | 18 | 360千瓦 | H100 | RD002 |
RD002E |
RD002A |
气体 |
| 40千瓦 | 10 | 360千瓦 | H100 | RD003 |
RD003E |
RD003A |
气体 |
| 130千瓦 | 108 | 7兆瓦 | GB200 | RD020 |
RD020E | RD020A | 液体 + 空气 |
| 130千瓦 | 18 | 1.2兆瓦 | GB200 | RD021 |
RD021E | RD021A | 液体 + 空气 |
| 58千瓦 | 10 | 500千瓦 | B200 | RD024 |
RD024E | RD024A | 气体 |
| 142千瓦 | 40 | 2.5兆瓦 | GB300 | RD027 |
RD027E | RD027A | 液体 + 空气 |
| 142千瓦 | 64 | 5兆瓦 | GB300 | RD028 |
RD028E | RD028A | 液体 + 空气 |
| 142千瓦 | 80 | 5兆瓦 | GB300 | RD029 |
RD029E | RD029A | 液体 + 空气 |
| 142千瓦 | 128 | 10兆瓦 | GB300 | RD030 |
RD030E | RD030A | 液体 + 空气 |
Frequently asked questions
What workloads is Vertiv™ 360AI optimized for?
Vertiv™ 360AI is designed for high-density AI workloads, including NVIDIA Blackwell DGX SuperPOD deployments. It supports up to 2.5 MW IT load with advanced power and cooling configurations.
Can I customize the reference design?
Yes. The design supports multiple thermal configurations (chilled water, direct expansion with refrigerant, and packaged systems) and modular power components, allowing customization based on site requirements and workload demands.
What cooling options are available?
Three thermal options are available:
Option 1: Chilled water with liquid-to-liquid CDU (XDU1350, CW205, FF3135)
Option 2: Direct expansion with refrigerant CDU (XDM300, DA265, MCV430)
Option 3: Packaged air and liquid cooling with CDU (DH400, XDU1350)
Each option supports N+1 redundancy and is designed to maintain optimal fluid temperatures around 23.8°C (75°F).
How does the AI Design Selector work?
It maps your workload to scalable power and cooling configurations, including:
- 4x 1000kVA UPS units (4-to-make-3 redundancy)
- 2x 240kVA UPS for facility systems (2N configuration)
- 8x 500kVA PDUs for rack-level distribution
- 400A busbars and 33kVA DC power shelves per rack
