Vertiv 360AI High Density Reference Design #006: 2x 1.28 MW NVIDIA Blackwell DGX SuperPODs
Vertiv
A three-page datasheet for a 2.56 MW AI hall built from two 1.28 MW SuperPODs, using four-to-make-three UPS redundancy and N+1 cooling.
On this page
Overview
This three-page datasheet (2025) describes a complete 2.56 MW hall built from two 1.28 MW SuperPODs. Total IT load is 2 x 1,280 kW = 2,560 kW. The room holds 32 compute racks at 73 kW, 16 network/storage racks at 14 kW and 4 cabling racks, 52 in all. Heat splits about 78/22 between direct-to-chip liquid and perimeter air, which is 992 kW liquid and 288 kW air per pod.
On the power side it uses a four-to-make-three (distributed redundant, 4N/3) block: 4 IT UPS of 1,200 kVA serve both pods, so three of the four installed modules carry the load. Cooling is N+1 and the cooling equipment (CDUs and CRAHs) has its own 240 kVA 2N UPS. The distribution chain is dual utility and generator feeds, two 4,000 A main switchboards, four 600 kVA PDUs per pod, 400 A overhead busways (8 per pod) and, in the rack, six OCP DC power shelves of 33 kVA each.
The design rests on Vertiv's six principles: co-design power and cooling, align clusters to capacity blocks, weigh TCO against redundancy and blast radius, buffer workload surges, plan hybrid liquid/air, and design for future density. The document expects AI rack density to grow toward 500 kW per rack. It is a vendor datasheet; its values are indicative and should be recomputed.
Key points
- Total IT load 2,560 kW = 2 pods x 1,280 kW.
- 32 racks at 73 kW + 16 racks at 14 kW + 4 cabling racks (52 racks).
- Per pod 992 kW liquid (78%) / 288 kW air (22%).
- Power is 4-to-make-3 (4 x 1,200 kVA IT UPS for 2 pods); cooling is N+1.
- Cooling UPS is 240 kVA, 2N, feeding both CDUs and CRAHs.
- Busways 400 A (8 per pod), tap-off boxes 60 A (2 per rack per busway); 6 x 33 kVA OCP DC power shelves per rack.
- CDU rated 1,350 kW (one per half-pod, 4 total); perimeter CRAH in the 150 kW class.
- Vertiv expects AI rack density to grow toward 500 kW per rack.
Using it in design
- Seed an AI-hall wizard template with: compute rack 73 kW, support rack 14 kW, liquid fraction 0.78, power 4N/3, cooling N+1, mechanical UPS 2N.
- Make it a rule that when liquid fraction is above zero, CDUs and pumps sit on UPS; flag designs that put CDUs on generator only.
- Use the 4-to-make-3 block to show how 3 x 1,200 kVA of usable capacity comes from 4 installed modules, each loaded to at most 75% normally.
Monthly newsletter · free
Keep your reading list current
Once a month: new library sources, changes to Turkish regulations and upcoming data centre events.
SoSo CAD
Review your power design in SoSo CAD
Inspect supply paths, equipment loads and UPS failure scenarios in one model.
Request a demo
Who should read it
Electrical and mechanical engineers designing AI halls in the 1-3 MW range.
Related standards
- Uptime Tier Standard: Topology
- TIA-942
- EN 50600-2-2
- OCP Open Rack V3
- ASHRAE TC9.9 Liquid Cooling Guidelines
Publication details
- Authors
- Vertiv
- Publisher
- Vertiv
- Year
- 2025
- Document number
- AI Reference Design #006
- Document language
- English
- Source type
- Manufacturer
- Access
- Free to read
- Licence
- Copyright retained; citation only — the publisher’s terms apply to the original document
Citation
Vertiv (2025). Vertiv 360AI High Density Reference Design #006: 2x 1.28 MW NVIDIA Blackwell DGX SuperPODs. Vertiv. AI Reference Design #006. https://www.vertiv.com/49cfcc/globalassets/documents/ai-hub-reference-design/vertiv-360ai-reference-design-006.pdf
Events on this subject
Browse the events calendarAll content and rights belong to the respective authors and publishers. SoSo CAD provides only the citation, its own summary and a link to the source, and is not responsible for the sources' content.