Case studies written as configurations, not logos.

Liquid cooling went from 15% of AI server shipments in 2024 to 54% in 2025 and a projected 76% in 2026 — and every one of those loops has to connect and disconnect dry, thousands of times per rack. The three delivery packages below are representative configurations: the hardware, the volumes and the timelines we actually ship, with customer identities kept out of it.

Case Studies

From GPU racks to battery packs.

Each configuration carries the same commercial spine — MOQ from 50 pieces, standard sets dispatched in 15 days, every batch helium-verified with its report in the carton — applied to three very different loops.

GPU server rack with blind-mate quick disconnects and a supply manifold feeding liquid-cooled trays

AI Data Centers · Retrofit

GPU rack manifold retrofit

A rack in the NVL72 class carries roughly 100 kW of cold-plate heat load, so tray drops convert to blind-mate UQDB04 pairs fed by an 8-port supply/return manifold pair per 42U rack. Trays pull hot without draining the loop — each disconnect loses at most 0.025 ml, the OCP dash-04 limit.

Rack class~100 kW cold-plate load
HardwareUQDB04 + 8-port manifolds
Loss per cycle≤ 0.025 ml

UQDB sets

Warm-water liquid cooling circuit with hand-push quick disconnects inside a supercomputer cabinet

HPC · Warm-Water Loop

Supercomputer warm-water circuit

Efficiency leaders on the Green500 now run 60–70 GFLOPS per watt on direct liquid cooling, which puts hand-serviceable ports on every blade row. UQD06 hand-push sets — the 3.0 GPM flow class — line the service side, with EPDM seals rated -40 to +120°C for glycol loops that never run cold enough to matter and never run hot enough to fail.

Flow classUQD06 · 3.0 GPM
Efficiency context60–70 GFLOPS/watt
SealsEPDM · -40..+120°C

HPC industry page

Battery energy storage container with pack-level liquid cooling manifolds and quick disconnect couplings

Energy Storage · Pack Line

BESS pack assembly line

Stationary BESS liquid cooling is growing at a 23.1% CAGR, and keeping cell-to-cell spread within ≤3°C only works if the loop can be opened safely at pack level. UQ04 hand-push sets at the pack manifold — 20 bar working, 80 bar burst — let a pack leave the line and come back for service dry.

Market context23.1% CAGR
Thermal target≤3°C cell spread
Pressure class20 bar · 80 bar burst

Energy storage page

Packed DryLink UQD coupling sets in foam-protected export cartons ready for dispatch

Delivery

What actually leaves the factory.

Every configuration above ends the same way: standard coupling sets dispatch in 15 days at MOQ 50, each set packed at 10×10×5 cm and about 0.2 kg, foam-protected and serial-labelled. Custom manifolds run 30–45 days after the drawing freezes, tray-packed with port caps fitted. Terms are FOB, CIF — insured at 110% of invoice value under ICC(C) clauses — or DDP to your door.

15 days — standard UQDB / UQ04 dispatch from order confirmation
10×10×5 cm · 0.2 kg — per-set export packing, serial-labelled
FOB / CIF / DDP — your forwarder collects, or we book the carriage

Representative Delivery

A configuration, stated the way we state all of them.

No testimonials, no invented customers — the delivery itself is the claim, and the batch report is the proof.

“Warm-water HPC service loop: 96 UQD06 hand-push sets across twelve blade rows, two 6-port service manifolds per cabinet, EPDM seals throughout — first articles approved inside the 15-day window, vacuum-decay and flow records packed carton by carton.”

REPRESENTATIVE CONFIGURATION · UQD06 + 6-PORT MANIFOLDS · MOQ 50 · 15 DAYS

Your loop looks like one of these — or like none of them.

Either way the quote path is identical: send the drawing or the photos, get a bore-to-bore answer within 48 hours.

Get a Quote Email Engineering