Who makes quick disconnects for AI data centers
Quick disconnects (UQD / UQDB)
Blind-mate connectors (UQD or UQDB) that let a compute tray or cold plate be connected to or disconnected from the coolant manifold without draining or spilling the loop.
Quick disconnects let a technician pull a single compute tray for service while the rest of the rack's coolant loop stays pressurized and sealed — a dripless, blind-mate connection at each branch off the manifold. Because a leak or a failed seal at this joint can damage the electronics it is meant to cool, the connector is qualified to tight tolerances against the specific manifold and cold plate it mates with, and reliability under repeated connect/disconnect cycles matters as much as flow rate. As liquid cooling moved from a minority of racks to the default configuration for AI accelerator trays, the number of these connectors per rack scaled with GPU count, and blind, dripless mating became a baseline requirement rather than a premium feature. Investors should watch qualification cycles at the connector level: a design that passes reference-design testing with a given GPU platform tends to be locked in for that platform's life, and few vendors carry the combination of sealing performance and manufacturing scale to serve hyperscale volumes.
AI delta: Liquid cooling moving from optional to default on AI accelerator trays scaled the number of blind-mate quick disconnects per rack with GPU count.
Companies on this part (7)
| Company | Role | Ownership | Ticker | Price | 1d | Confidence | Note |
|---|---|---|---|---|---|---|---|
| CPC/Colder · unit of Dover | manufacturer | subsidiary | DOV* | inferred | |||
| Eaton | manufacturer | public | ETN | 366 USD | -1.7% | inferred | |
| Fositek | manufacturer | public | 6805.TW | 2335 TWD | -2.5% | reported | |
| Parker Hannifin | manufacturer | public | PH | 881 USD | -1.4% | inferred | Hannifin |
| Danfoss | manufacturer | private | disclosed | · source | |||
| Stäubli | manufacturer | private | inferred | ||||
| Swagelok | manufacturer | private | disclosed | · source |
Who competes here
The same table read the other way: every company above competes for quick disconnects sockets. Confidence tags matter — an inferred placement is a lead, not a confirmed rivalry.
Seen in real machines
quick disconnects is named in 2 of the rack teardowns and supercomputer breakdowns on this site. Each line below is a piece of that machine. The companies shown are the ones named on that line, not necessarily the maker of this part.
- Nvidia GB300 NVL72 · rack teardown
- Quick disconnects — Stäubli, Parker Hannifin, Danfoss, Fositek
- Fugaku · supercomputer
Nearby parts
- Previous in layer: Rack manifolds
- Next in layer: Hoses, tubing, fittings, valves
- Layer: Layer 5 — Liquid cooling (IT-side: chip to CDU)
From the blog
- Part of the day: Quick disconnects 2026-10-06
Watch
Independent videos — no affiliation; nothing loads from YouTube until you click.
Latest feed
- Sponsored: GF completes LiquidCore portfolio with intelligent Multi Control Valve to maximize direct liquid cooling efficiency — Data Center Dynamics, 2026-10-24GF has launched a Multi Control Valve to finalize its LiquidCore system, enhancing the active intelligence layer for direct liquid cooling.
- Interview: TI sees liquid cooling shift to tray-level control in AI data centers — DigiTimes, 2026-10-23Texas Instruments stated that the trend in AI data centers is moving toward granular, tray-level thermal control rather than relying solely on centralized CDU management.
- [AI Infrastructure Insight] How AI infrastructure will be designed for the future — newsroom:sk-hynix, 2026-10-22The design of future AI infrastructure will require integrating advanced compute, cooling, and power elements from various parts of the supply chain.
- Global Tek's new Thailand plant enters trial production in 3Q26, targeting AI liquid cooling and aerospace opportunities — DigiTimes, 2026-10-20Global Tek Industrial opened its large Thai facility, which will expand production to include components for AI liquid cooling alongside its existing aerospace and automotive work.
- Warm-water ready — Data Center Dynamics, 2026-10-15The report compares the performance and efficiency of two-phase versus single-phase direct-to-chip cooling solutions specifically on NVIDIA B200 server hardware.
- MiTAC spotlights liquid cooling, open platforms for AI infrastructure — DigiTimes, 2026-10-14MiTAC showcased liquid cooling solutions and open platforms designed for advanced AI infrastructure requirements.
- Qisda's Aewin targets 200kW liquid-cooled AI servers — DigiTimes, 2026-10-08Qisda's Aewin brand is targeting the market with 200kW liquid-cooled AI servers designed for high-density data center deployments.
- DIGITIMES Insight: AI data demand surges as power and cooling become new bottlenecks — DigiTimes, 2026-10-04Increasing AI data center demand is creating significant logistical and infrastructure bottlenecks specifically regarding power availability and cooling capacity.
- NVIDIA Announces DSX Ready Qualification Program for Data Center Power and Cooling Hardware — ServeTheHome, 2026-09-22NVIDIA launched the DSX Ready program, a qualification standard for data center power and cooling hardware used in AI data center construction.
- FCT Introduces Liquid Leak Detection Flex Assembly for AI and HPC Systems — HPCwire, 2026-09-21FCT released a Liquid Leak Detection Flex Assembly, a custom sensor for liquid-cooled AI and HPC systems, covering CDUs, manifolds, and quick disconnects.
Questions about quick disconnects
- What are quick disconnects?
- Blind-mate connectors (UQD or UQDB) that let a compute tray or cold plate be connected to or disconnected from the coolant manifold without draining or spilling the loop.
- Which companies make quick disconnects for AI data centers?
- This catalog lists 7 companies on Quick disconnects, 4 of them with listed equity: CPC/Colder, Eaton, Fositek, Parker Hannifin, Danfoss, Stäubli, Swagelok.
- How does AI change demand for quick disconnects?
- Liquid cooling moving from optional to default on AI accelerator trays scaled the number of blind-mate quick disconnects per rack with GPU count.
- How reliable are these quick disconnects supplier placements?
- The 7 placements on this part break down as 2 disclosed, 1 reported, 4 inferred. "disclosed" means the company or a filing says so; "reported" comes from trade press or a reference-design list; "inferred" is a taxonomy lead with no citation yet.
Sources: placements without a source URL come from the taxonomy seed and are tagged accordingly. How this data is built.· Markdown · JSON