I specify the fluid between the chiller and the rack — secondary coolants & heat transfer fluids, -130 °C to +350 °C. Data centre, process, cold chain.
The fluid is decided last and blamed first.
Rising pump energy, fouling, a cracked plate, conductivity drift in a live rack — almost always decided months earlier, at selection.
I work the fluid side: −130 °C to +350 °C. Loop questions welcome.
@b_radford@GoodisonJason Service-model point is right — liquid racks change O&M. Beyond leak sensors and re-torques, name who trends coolant chemistry after day-one sample; drip risk and fluid health are different owners.
@Bytes_Equity Aeroflex LC products into AI DCs maps the SFN/hose layer well. After skids ship, who owns secondary chemistry at the CDU–rack interface — supplier or site — still decides handover quality.
@TheAijazHussain Localized power plant wrapped around dense silicon — thermal is half the plant. Secondary loop ownership (flush-fill + chemistry sign-off) is the quiet gate before the silicon can stay at TDP.
@J5jayfive Power + liquid cooling as the AI compounder is fair. Conversion still hinges on commissioning: named owner for secondary fluid QA after CDU install turns kits into soak-ready clusters.
@while_int HVAC giants into DC liquid cooling is the channel shift. After CDUs land, who trends inhibitor/particulate on the secondary — AC OEM, CDU vendor, or site — still decides year-one boring uptime.
@DCX_LiquidCool@ForbesPolska Forbes private ranking is a strong milestone for a liquid-cooling OEM. As ECDU/FDU scale, the quiet differentiator remains who owns fluid QA after first fill — hardware rank ≠ ops ownership.
@AccelsiusATX Two-phase D2C for AI factory heat is the right constraint frame. At scale, facility secondary still needs a named owner for flush-fill QA — socket cooling ≠ site loop chemistry signed.
@BharatTechIND 52 MW AI-ready + advanced liquid cooling in Chennai is the build headline. Commissioning gate is still who owns secondary flush-fill and chemistry before first racks soak — plant ≠ fluid-ready.
@JulesNauy Platform path (DLC vs immersion) steers supplier capex hard. Whatever wins, site still needs a named owner for fluid chemistry after factory seal meets facility water — path choice ≠ ops ownership.
@andyrapista PH colo MW + cable routes is the capacity story. For AI density asks, worth clarifying liquid-ready secondary loops and who signs first-fill chemistry — air halls and LC halls commission differently.
@Nodus42 Carrier order spike maps the HVAC→DC cooling channel. After skids land, who owns secondary loop QA — OEM, MEP, or site — still decides whether backlog converts to soak-ready halls.
@pequityresearch Retrofit liquid adoption is where schedules slip. Capex models the CDU; ops models who signs secondary chemistry and particulate after first fill — that name decides year-one uptime.
@CoreWeave 10-second rack shutoff is the right leak failsafe. Still need a named owner for secondary flush-fill and chemistry before soak — Valvey protects the event; fluid QA prevents the slow death.
@ChinaMobile_X WUE 0.4 via liquid + reclaimed water is a strong ledger. Closed-loop secondary chemistry and particulate still decide whether that WUE holds after year-one fill — water meter ≠ fluid health meter.
@VertivAP Free-cooling + low-GWP trim is the facility-side efficiency story. Secondary loop still needs inhibitor/particulate trending after first fill — energy wins don't replace fluid-QA ownership.
@VertivAP Power + liquid + controls as one stack is the right JPCDC frame. Commissioning still needs a named owner for secondary flush-fill and chemistry at the CDU — converge hardware, then converge who signs fluid readiness.
@_NickLynch_@JayG2MAX ASHRAE facility-water vs cold-plate loop split is the ownership map. Naming one coolant-health owner on each side — before racks soak — is what keeps mixed-vendor halls from becoming a blame loop.
@JayG2MAX@_NickLynch_ Agree — split across CDU manual / fit-out / lease is how 3am calls get born. One signed page naming fluid-QA owner at the CDU outlet before soak beats three unread PDFs.
Cx can sign one chemistry sample at the CDU outlet on handover day. Day two still needs a named owner for the trending log — inhibitor, particulate, ΔP. How often is that name actually written into the O&M contract before racks soak?
@9cloudai Parallel handover + equipment install compresses the schedule — and compresses fluid QA. Who signs secondary chemistry before soak when the machine room and fit-out land in the same window?