Case studies — Hydroniq Technologies

— CASE 01

BFSI · Tier-1 Indian bank

CHENNAI · 2026 Q1
SOLUTION · DLC + RDHX

Inference-cluster retrofit inside an active BFSI hall.

Problem

Existing Tier-III hall designed for 7 kW racks needed to host a new fraud-detection inference cluster averaging 38 kW per rack. No appetite for taking the room offline. No appetite for full immersion. Three-month window.

Solution deployed

Direct-to-chip cold plates on the GPU servers, paired with in-row CDUs and Motivair ChilledDoor rear-door heat exchangers on adjacent racks to absorb residual air load. New 600 kW TCS loop tied into the existing chilled-water plant. Castrol ON DLC fluid.

Outcome

Cluster live four weeks ahead of plan. Surrounding rows remained in production throughout. Acoustic levels in the hall actually dropped after retrofit due to lower fan duty.

Rack density

42kW

PUE before

1.62

PUE after

1.18

Downtime

0hrs

— CASE 02

IT services · GCC operator

BENGALURU · 2026 Q2
SOLUTION · SINGLE-PHASE IMMERSION

Greenfield AI training pod for an internal LLM programme.

Problem

Global IT services firm building an India-based training pod for an internal foundation model. Required 1.5 MW of GPU compute in a 220 m² envelope with PUE under 1.2. Conventional air design needed nearly 4× the floor plate.

Solution deployed

Six Submer SmartPod immersion tanks, single-phase, populated with H100-class servers from a major OEM on the Submer compatibility list. Submer SmartCoolant fluid, supplied and commissioned in-country. Vertiv Liebert XDU loop with 2N pumps and dry-cooler heat rejection on the roof.

Outcome

1.5 MW of training capacity in 210 m². No CRAC units on the whitespace. Complete elimination of hot-aisle work for the ops team. Annual cooling energy benchmarked at ~30% of the equivalent air-cooled design.

Rack density

120kW

PUE achieved

1.08

Floor saved

72%

Total IT load

1.5MW

— CASE 03

Hyperscale · Cloud operator

HYDERABAD · 2026 Q3
SOLUTION · DLC AT SCALE

Multi-megawatt direct-to-chip rollout across a hyperscale region.

Problem

Cloud operator standing up a new AI region needed an on-the-ground partner to handle CDU procurement, fluid logistics, and commissioning support across the first three buildings.

Solution deployed

Programme-managed supply of CoolIT CDUs across multiple halls, in-country fluid stocking with Shell, and a commissioning team embedded with the operator's facilities group through go-live.

Outcome

Detailed write-up available under NDA after the first building goes GA. Topline metrics tracking ahead of design PUE.

Rack density

130kW

Target PUE

1.10

Halls

3

Total IT load

~12MW

Selected deployments.

Inference-cluster retrofit inside an active BFSI hall.

Greenfield AI training pod for an internal LLM programme.

Multi-megawatt direct-to-chip rollout across a hyperscale region.

We can walk you through any of these in person.