

DRACO 2.5 MW Facility CDU
2.5 MW · Up to 4,000 L/min · Liquid-to-liquid facility CDU · PG25 / treated water · Facility CDU (floor-standing, N+1)
Key Specifications
See full specs ↓“RDP delivered and installed our edge AI pods across 6 sites with predictable INR pricing and onsite SLA.” — [customer / sector, to confirm]


Designed, built and supported in India — sovereign by design
Your AI factory on sovereign Indian infrastructure: data residency under DPDP, MeitY-recognised, ISO 27001 / SOC 2 deployment paths, and procurement on GeM.
Overview
The DRACO 2.5 MW Facility CDU is RDP’s flagship coolant distribution unit — 2.5 MW of heat removal that cools a full row-zone or a data-hall block of dense GPU racks. It is the facility-scale thermal backbone that makes the largest on-premises AI deployments possible, in INR, on a GST invoice.
Engineered to match RDP’s rack-scale systems, superclusters and supercomputers, it isolates the rack coolant loops from facility water, controls flow, temperature and pressure, filters the coolant, and distributes it across many racks with N+1 pumping — delivered sized, plumbed and commissioned with the compute it cools, with leak detection and monitoring built in.
Key highlights
- 2.5 MW · Up to 4,000 L/min — facility-scale heat removal for Up to 24 dense racks / a row-zone.
- Liquid-to-liquid facility CDU — isolates rack coolant from facility water; controls flow, temperature and pressure with filtration at scale.
- PG25 / treated water — corrosion-safe coolant chemistry for reliable long-term operation.
- N+1 redundant pumps — continuous cooling during maintenance and failures.
- Leak detection & monitoring — sensors and BMS/DCIM integration for safe, observable operation at scale.
- Efficient, low PUE — liquid removes heat far more efficiently than air across the hall.
- Validated with RDP compute — delivered as part of a tested, cooled data hall.
- Make-in-India OEM — predictable INR pricing, GST tax invoice, pan-India onsite support, GeM-procurable.
Where it fits
- Data-hall cooling (primary): distributes coolant to direct-to-chip loops across Up to 24 dense racks / a row-zone.
- Superclusters & supercomputers: the cooling backbone for halls of NVL72-class racks.
- New liquid-cooled facilities: the facility CDU layer between the racks and the building’s water loop.
- Scale-up cooling: add CDUs as the data hall grows.
How it works
The facility CDU forms a closed technology-cooling-system (TCS) loop to many racks, rejecting heat to the building water loop through large heat exchangers. It controls flow, supply temperature and pressure, filters and treats the coolant, and provides N+1 pumping for continuous operation. RDP sizes the CDU, loops and connections to your hall density and facility water conditions. Honest note: real cooling capacity depends on inlet water temperature, flow and hall layout — we size and validate it for your facility.
Industry use cases
- AI data centres — cool halls of dense GPU and NVL racks at full clocks.
- Government & national labs — sovereign HPC/AI cooling at facility scale.
- Neocloud / hyperscale operators — facility-scale density and low PUE.
- Large enterprise & energy — the largest HPC + AI thermal deployments.
- Colocation — liquid-ready density across shared halls.
- National research programmes — cooling for the largest clusters.
Performance — and how to be sure
We don’t publish inflated peak numbers. The honest picture: 2.5 MW at Up to 4,000 L/min cools a row-zone of dense AI racks within thermal limits given suitable facility water. Want certainty? Request a thermal assessment of your hall and facility before you buy; we’ll size the CDU and confirm the cooling capacity at your conditions.
Series & scale path
- DRACO (flagship thermal tier) — this, the largest in the line.
- Capacity ladder: rear-door HX → in-rack/in-row CDU → row CDU → facility CDU (MW-scale).
- When to step up: add facility CDUs as the hall grows — talk to an architect about the loop and redundancy.
Liquid vs air — the efficiency case
At 700 W–1 kW+ per GPU, air cooling hits its limit; liquid removes heat far more efficiently across the hall, enabling higher density, full clocks and lower PUE. The result is more compute per facility and lower energy cost. RDP pricing is fixed in INR with a GST input-credit-eligible invoice — ask for a PUE and density comparison for your facility.
Software & integration
Integrates with your facility: BMS/DCIM monitoring, leak detection, flow/temperature/pressure telemetry, N+1-pump control and alarms. Designed to match RDP rack-scale systems, superclusters and supercomputers and standard facility water loops.
Power, cooling & facility integration
A Facility CDU (floor-standing, N+1) unit — plan facility water (supply/return temperature and flow), power for pumps/controls, and floor integration with redundancy. (Exact capacity, flow, pressure-drop and connection figures confirmed in the thermal design.) N+1 pumps and full monitoring.
Deployment, warranty & support
- Made to order, sized, plumbed and commissioned in India with the compute it cools; project timeline confirmed at quote.
- In the box: facility CDU, pumps, heat exchangers, fittings, sensors and controls; facility connection scoped at quote.
- Onsite warranty + AMC with pan-India coverage and an RMA/escalation path (exact terms confirmed at quote).
Why RDP
14 years of Make-in-India infrastructure and 300,000+ devices shipped. Indian OEM, INR pricing, GST tax invoice, pan-India onsite engineers, GeM availability, and DPDP / sovereign-AI-ready deployment.
Buy with confidence
This is facility-scale AI cooling, made to order — talk to an RDP solution architect, get a thermal assessment and CDU sizing for your hall and facility, and a PUE comparison before you commit. Request a quote to begin.
Specifications
| Series | DRACO |
| Form Factor | Floor-standing |
| Cooling Capacity | 2.5 MW |
| Flow Rate | Up to 4,000 L/min |
| Type | Liquid-to-liquid facility CDU |
| Coolant / Fluid | PG25 / treated water |
| Racks Supported | Up to 24 dense racks / a row-zone |
| Model fit | Facility-scale heat rejection |
Why RDP GPU Mart
- ✓ Make in India OEM — Hyderabad facility, 14 years, 300,000+ devices shipped.
- ✓ Sovereign-ready: India data residency (DPDP), MeitY-recognised, ISO 27001 / SOC 2 paths.
- ✓ INR-transparent: GST invoice, CGST/SGST or IGST, pan-India onsite SLA.
- ✓ Available on GeM for government and PSU procurement.
FAQ
Is GST invoicing available?
Yes — GST invoice, CGST+SGST or IGST by billing state, eligible for input credit.
Do you deliver and install pan-India?
Yes — pan-India delivery with onsite installation and a 3-year onsite SLA.
What warranty and support is included?
3-year pan-India onsite SLA with AMC and flexible financing options.
Can this be configured to my workload?
Yes — talk to an RDP solutions architect for a custom build or multi-node cluster.
Compare the range
Other Liquid Cooling & Thermal in this line
Swipe to compare
| 40 kW Rear-Door H… | 100 kW In-Row CDU | Direct-Liquid Coo… | 50 kW Direct-Liqu… | |
|---|---|---|---|---|
| GPUs | — | — | — | — |
| GPU memory | — | — | — | — |
| Model fit | Air-assist rack cooling | Liquid-cooled GPU racks | Direct-to-chip liquid cooling | Direct-to-chip liquid cooling |
| Networking | — | — | — | — |
| Chassis | — | — | — | — |
| Price | Request a Quote | On request | Request a Quote | Request a Quote |
| View | Quote | View | View |
Build the full stack
Pair it with








Designing a GPU cluster, not just one server?
Talk to an RDP solutions architect about the full fabric — networking, storage, rack and power.
*Pan-India delivery and onsite installation are subject to location serviceability; standard SLA terms apply. Specifications indicative; final configuration confirmed on quote.