See more
Paired to perform.
I want to
retrofit an existing
data center
I want to
build a new
data center
Tap to see the path to full-stack accelerated computing platforms for generative AI workloads.
Ready for what's next?
See more
The Vertiv and NVIDIA co-development partnership enables customers to transform traditional data center architectures into AI Factories capable
of powering AI applications across the enterprise.
Working together, we’re setting a
future-forward roadmap for
technical co-development and enabling the deployment of accelerated computing at scale.
Solving the Complexities
of the AI Revolution
Home
In partnership with
360kW cluster
perimeter
air cooled
20kW
500kW cluster
perimeter
air cooled
20kW
360kW cluster
perimeter
air cooled
40kW
360kW cluster
RDHx
air cooled
40kW
500kW cluster
RDHx
air cooled
58kW
1 POD
liquid & perimeter
73kW
340kW cluster
liquid & RDHx
75kW
1.2 MW
Design
1 POD
1.2 MW
Design
1 POD
2.3 MW
Design
2 PODS
3.5 MW
Design
3 PODS
4.6 MW
Design
4 PODS
7.0 MW
Design
6 PODS
5.0 MW
CW Design
2 POD
Up to
80kW
per rack
Up to
132kW
per rack
Up to
300kW
per rack
Up to
80kW
per rack
Up to
132kW
per rack
Up to
300kW
per rack
Home
Retrofit
Building your
AI Factory.
Scan to get the full reference design.
Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Lorem ipsum dolor sit amet,
Total IT Load 358kW
AI Reference Design #017
10
Racks
360kW Cluster at
up to 41kW per with rear-door heat exchangers (RDHX)
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for a 2.56MW solution with four to make three redundancy,
using two NVIDIA Blackwell DGX SuperPODs.
32 racks at 73kW, 16 racks at 14kW,
4 cabling racks
Total IT Load: 2,560kW (1280kW per POD)
AI Reference Design #006
52
Racks
2x 1.28MW NVIDIA Blackwell
DGX SuperPODs
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for
a 500kW solution with N configuration,
using NVIDIA DGX B200 System.
7 racks at 58kW, 1 rack at 43kW,
2 racks at 20kW
Total IT Load 484kW
AI Reference Design #024
10
Racks
500kW Cluster at up to
58kW per rack with rear-door
heat exchangers (RDHX)
Representation
Representation
Scan to get the full reference design.
Designed for a 360kW solution with 2N redundancy,
this full data center environment uses the
NVIDIA DGX H100 System.
7 racks at 41kW, 1 rack at 30kW,
2 racks at 20kW
Total IT Load 358kW
AI Reference Design #004
10
Racks
360kW Cluster at
up to 41kW per with rear-door heat exchangers (RDHX)
Representation
Representation
Scan to get the full reference design.
Designed for a 360kW solution with 2N redundancy,
this full data center environment uses the
NVIDIA DGX H100 System.
Rack Density: 7 racks at 41kW, 1 rack at 30kW, 2 racks at 20kW
Total IT Load 358kW
AI Reference Design #003
10
Racks
360kW Cluster at
up to 41kW per rack air cooled
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for
a 500kW solution with N configuration,
using NVIDIA DGX B200 System.
16 racks at 28.6kW, 2 racks at 20kW
Total IT Load 484kW
AI Reference Design #022
18
Racks
500kW Cluster at 29kW
per rack air cooled
Representation
Representation
Scan to get the full reference design.
Featuring the NVIDIA DGX H100 System,
this full data center environment designed for
a 360kW solution with 2N redundancy.
16 racks at 20.5KW per rack, 2 racks at 20kW
Total IT Load 358kW
AI Reference Design #002
18
Racks
360kW Cluster at
20kW per rack air cooled
Representation
Representation
See more
A full data center environment designed for a
7MW solution with 4 to make 3 redundancy, using six
NVIDIA DGX SuperPOD™ systems utilizing the NVIDIA GB200 NVL72 platform. The latest end-to-end power and cooling solution with OCP-compliant components.
6x 1.2MW NVIDIA DGX SuperPODs™
48 racks at 130kW, 48 racks at 14kW
AI Reference Design #020
108
Racks
6,912kW (1,152kW per POD)
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for a 4.6MW solution with four to make three redundancy,
using four NVIDIA Blackwell DGX SuperPODs.
4x 1.15MW NVIDIA Blackwell DGX SuperPODs
Rack Density: 32 racks at 130kW, 32 racks at 14kW, 8 cabling racks
AI Reference Design #016
72
Racks
Total IT Load 4,608kW
(1152kW per POD)
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for
a 3.5MW solution with four to make three redundancy, using three NVIDIA Blackwell DGX SuperPODs.
3x 1.15MW NVIDIA Blackwell DGX SuperPOD
Rack Density: 24 racks at 130kW, 24 racks at 14kW, 6 cabling racks
AI Reference Design #015
54
Racks
Total IT Load 3,456kW
(1152kW per POD)
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for
a 2.3MW solution with four to make three redundancy, using two NVIDIA Blackwell DGX SuperPODs.
2x 1.15MW NVIDIA Blackwell DGX SuperPODs
Rack Density: 16 racks at 130kW, 16 racks at 14kW, 4 cabling racks
AI Reference Design #014
36
Racks
Total IT Load 2,304kW
(1152kW per POD)
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for a 1.13MW solution with four to make three redundancy, using one NVIDIA Blackwell DGX SuperPODs.
1x 1.13MW NVIDIA Blackwell DGX SuperPODs
Rack Density: 8 racks at 127kW, 8 racks at 14kW, 2 cabling racks
AI Reference Design #021
Total IT Load 1128kW per POD
18
Racks
Representation
Representation
Scan to get the full reference design.
A full data center designed for a 1.2 MW Nvidia
DGX GB200 NVL72 SuperPod.
1.2 MW NVIDIA Blackwell DGX SuperPODs
Rack Density: 16 racks at 132kW, 20 racks at 12kW
AI Reference Design #019
19
Racks
Total IT Load 1.2 MW
Representation
Representation
Scan to get the full reference design.
A full data center environment designed for a
5MW solution with four to make three redundancy,
using two NVIDIA Rubin Next DGX SuperPODs.
2x 2.5MW NVIDIA Rubin Next DGX SuperPODs
Rack Density: 16 racks at 269kW, 16 racks at 43kW, 4 cabling racks
AI Reference Design #300
36
Racks
Total IT Load 4,992kW
(2496kW per POD)
Representation
View Schematics
Representation
Find your solution.
In partnership with
High capacity
1.4MW solution
with supplemental perimeter
air cooling
High capacity 1.2MW solution with heat reuse
Liquid-to-liquid solution with supplemental perimeter
air cooling
High density
air-cooled
solution with
DX heat rejection
Liquid-to-liquid solution with heat reuse
Low cost
liquid-to-refrigerant solution
Air-cooled
retrofit
In-rack
liquid-to-liquid solution
High density
liquid-to-air solution
Large space applications
Medium space applications
Small space applications
Large space applications
Medium space applications
Small space applications
Home
New Build
Why Vertiv?
See more
For facilities that are unable to change existing
architectures and do not have chilled water available
on site, this solution offers a path to introducing liquid cooling into the existing space.
Liquid to air direct-to-chip retrofit.
Small HPC minimal retrofit
Heat removal
from room
Heat removal
from server
Brown field
88kW
Load per rack
No water.
No problem.
1
Rack(s)
88kW
Total Solution
Capacity
See more
Want a full liquid-to-liquid solution with a small footprint? This solution doesn’t need additional floor space for coolant distribution units and leverages existing air-cooling for the remaining portion of the heat load.
Liquid to air direct-to-chip retrofit.
Small HPC minimal retrofit for chilled water system
Heat removal
from room
Heat removal
from server
Brown field
100kW
Load per rack
100kW
Total Solution
Capacity
1
Rack(s)
Small Space.
Big Performance.
See more
Designed specifically for high density air-cooled servers, this retrofit solution provides room neutral cooling in the form of a rear door heat exchanger. Ideal for edge environments or AI inferencing applications.
High density air-cooled retrofit
Small HPC minimal retrofit
Heat removal
from room
Heat removal
from server
Brown field
80kW
Load per rack
80kW
Total Solution
Capacity
1
Rack(s)
Kick start
your edge
See more
Intended for retrofit of a data center with existing infrastructure and Vertiv cooling systems,
this budget-friendly solution can leverage existing
refrigerant piping and heat rejection
Liquid to refrigerant direct-to-chip
retrofit-optimized row.
Mid-sized HPC cost-optimized retrofit, chiller included
Heat removal
from room
Heat removal
from server
Brown field
100kW
Load per rack
400kW
Total Solution
Capacity
4
Rack(s)
Power that doesn’t heat up your budget.
See more
This solution maintains and optimizes the existing footprint while maintaining room neutrality by combining the use of direct-to-chip liquid cooling with a rear door heat exchanger.
Liquid to liquid direct-to-chip with heat reuse.
Mid-sized HPC with increased heat capture.
Heat removal
from room
Heat removal
from server
Brown field
Green field
100kW
Load per rack
400kW
Total Solution
Capacity
4
Rack(s)
Putting the fit
in retrofit.
See more
This solution retrofits existing air-cooled environments without disrupting existing workloads by incorporating
rear-door heat exchangers with refrigerant-based indoor split chillers. It also incorporates existing Vertiv™ Liebert® DSE cooling units and reuses the existing refrigerant piping and heat rejection. It doesn’t require chilled water loops on site.
High density air-cooled retrofit
Mid-sized HPC pragmatic retrofit for air-cooled computer rooms, chiller included.
Heat removal
from room
Heat removal
from server
Brown field
Green field
40kW
Load per rack
160kW
Total Solution
Capacity
4
Rack(s)
An upgrade
made easy.
See more
Intended to replace and optimize existing footprints for
high-density and AI applications. The solution combines
the use of liquid cooling direct-to-chip with air cooling
to cover the remaining capacity.
Liquid to liquid direct-to-chip for
early adoption.
Mid-sized HPC low complexity retrofit with air-cooling.
Heat removal
from room
Heat removal
from server
Brown field
100kW
Load per rack
500kW
Total Solution
Capacity
5
Rack(s)
More power.
Same space.
See more
Intended for large, new deployments that require heat capture for room neutrality, this solution combines the use of direct-to-chip liquid cooling with a rear door heat exchanger.
Liquid to liquid direct-to-chip with
heat reuse at scale
Large HPC preserving room neutrality.
Heat removal
from room
Heat removal
from server
Brown field
Green field
100kW
Load per rack
1200kW
Total Solution
Capacity
12
Rack(s)
Neutrality
made easy.
See more
Intended for large, new deployments, this solution combines the use of direct-to-chip liquid cooling with existing air cooling to cover the remaining capacity.
Liquid to liquid direct-to-chip at scale.
Large HPC building towards scale.
Heat removal
from room
Heat removal
from server
Brown field
Green field
100kW
Load per rack
1400kW
Total Solution
Capacity
14
Rack(s)
XL Peace of Mind.
Design for the future. AI powerdensity will rapidly grow toward 500kW per rack.
Design for a mix of liquid and air cooling,
which have a combined and interdependent
impact on the ability to remove heat.
Handle AI workload surges though system-level controls and power and cooling buffers.
Consider total cost of ownership, redundancy, and blast radius in AI power and cooling designs, and the tradeoffs among them.
Dial up power efficiencies. Eliminate stranded
power by aligning AI clusters to data center
‘capacity blocks.’
Design both power and cooling together to optimize AI infrastructure.
Vertiv can help you:
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
View Schematics
Thermal Management
Thermal Control
Heat Rejection and Reuse
Air Cooling
Liquid Cooling
Chip to Heat Reuse
View in 3D
Select a higlighted product for more info.
High-Density Data Center
Traditional
Data Center
Thermal Control
Heat Rejection and Reuse
Services
Air Cooling
Hybrid Cooling
Liquid Cooling
Home
See more
Liquid Cooling Services
Services
See more
Thermal
Services
Services
See more
Liebert® iCOM™
CWM Chilled Water
Plant Manager
Thermal Control
See more
Liebert®
iCOM™ Cooling
Unit Control
Thermal Control
See more
Liebert® iCOM™-S
Advanced Thermal
Monitoring and
Control Software
Thermal Control
See more
Thermal Control
See more
Vertiv™ Liebert® MC Microchannel Coil Condenser
Heat Rejection and Reuse
See more
Liebert® MCV High Density Microchannel Condenser
Cost: Up to 15% TCO improvement
Heat Rejection and Reuse
See more
Vertiv™ CoolLoop
Trim Cooler
Efficiency: Up to 70% energy savings
Space: Up to 40% increase in density
Heat Rejection and Reuse
See more
Vertiv™
Liebert® AFC
Free-Cooling Chiller
Efficiency: Up to 20% energy savings
Heat Rejection and Reuse
See more
Heat Rejection
and Reuse
See more
Vertiv™ CoolPhase
Flex Packaged
Cooling System
Speed: Up to 50% faster deployments
Hybrid Cooling
See more
Liebert® CWA Thermal Wall
Air Cooling
See more
Liebert® CW Chilled Water Cooling System
Air Cooling
See more
Vertiv™ Liebert®
DSE Packaged
Free-Cooling Solution
Air Cooling
See more
Liebert® DS
Direct Expansion
Cooling System
Air Cooling
See more
Liebert® DSE
Free Cooling Economization System
Cost: Up to 15% TCO improvement
Air Cooling
See more
Air Cooling
See more
Vertiv™
CoolCenter
Immersion Systems
Liquid Cooling - Coming Soon
See more
Vertiv™ CoolChip
Fluid Network
In-Rack Manifolds
Liquid Cooling
See more
Liebert®
DCD Rear Door
Heat Exchangers
Liquid Cooling
See more
Vertiv™ CoolPhase
CDU Coolant
Distribution Units
Cost: Up to 15% TCO improvement
Liquid Cooling
See more
Vertiv™ CoolChip
CDU Coolant
Distribution Unit
Efficiency: Up to 15% Energy Savings
Liquid Cooling
See more
Liquid Cooling
See more
Vertiv™ Unify
Thermal Control
