Published 2026 market sizing for coolant distribution units (CDUs) ranges from USD 1.03 billion [S7] to USD 11.4 billion [S4] in the same calendar year, a spread of more than 10x driven entirely by how each analyst defines the addressable market, from factory-floor process coolant skids to hyperscale data center liquid-cooling racks.
The narrowest definition (Fact MR, 2026) tracks general industrial CDUs at USD 1.03 billion rising to USD 6.98 billion by 2036 [S7]; mid-scope industrial/electronics CDUs land at USD 1.50-2.54 billion in 2026 depending on inclusion of heat exchangers and secondary-loop controls [S1][S3]; the broadest data-center-only slice (Stratview, 2026) hits USD 11.4 billion, up 36.8% year-over-year against 2025's USD 8.4 billion base [S4].
Scope-driven sizing: industrial CDU vs. data center CDU vs. liquid cooling
Industrial coolant distribution units serving metalworking, plastics, HVAC and process plant applications are sized at USD 1.8 billion in 2026 with a 7.2% CAGR through 2033 [S2], while a parallel estimate puts the same industrial/electronics pool at USD 2.54 billion in 2026 growing at 14.3% CAGR through 2034 [S1]. Data center CDU-only reporting sits at USD 1.10 billion (IntelMarketR, 2026) growing to USD 3.90 billion by 2034 at 21.8% CAGR [S5], and a wider Stratview cut covering facility and rack-level CDUs lands at USD 11.4 billion in 2026 with a 35.3% CAGR to USD 70.5 billion in 2032 [S4]. The total data center liquid cooling market, which bundles CDUs, cold plates, immersion tanks, manifolds and coolant, is sized at USD 6.77 billion in 2026 by Mordor [S6] and USD 4.6 billion in 2025 by Market.us [S8].
Why rack density is the structural driver, not the headline number
Rack configurations exceeding 80 kW are now routine in AI server deployments, with installations approaching or exceeding 120 kW per rack, a thermal envelope that air handling cannot cost-effectively remove [S5]. This density shift, not a general server capex boom, is the physical reason liquid cooling is moving from a niche to a baseline requirement. Hyperscale operators are embedding coolant distribution unit architectures into greenfield campuses and large-scale retrofits rather than treating them as optional add-ons [S3].
The Polaris report ties this to a documented capex signal: UNCTAD reported announced foreign direct investment into data centers exceeding USD 270 billion in 2025 alone, with hyperscale campuses designing new builds around high-density rack cooling frameworks from day one [S3]. For comparison, the IEA estimates global electricity consumption by data centres, AI and cryptocurrency will exceed 1,000 TWh per year by 2026, equal to Japan's annual electricity use [S5], a baseline that frames the cooling-energy trade-off CDUs are now being judged against.
Segment split: FDU, direct-to-chip, hyperscale lead the growth lanes

By type, FDU (fluid distribution unit, the rack-level manifold/CDU hybrid) is the fastest-growing CDU segment through 2032 [S4]. By cooling architecture, direct-to-chip cooling is the fastest-growing sub-segment, followed by immersion cooling in second-tier deployments [S4][S5]. By end user, hyperscale operators lead adoption, with colocation and enterprise trailing but accelerating as rack densities in mid-tier facilities push past the 30-40 kW practical limit of air cooling [S4].
The facility loop versus secondary loop split is now a default architectural line: the facility loop circulates water from central plant infrastructure, while the CDU-managed secondary loop isolates and delivers conditioned coolant directly to IT equipment through heat exchangers [S3]. This split is what lets a campus hot-aisle layout survive a refrigerant change, and it is why a hydraulic power unit and pump skid can be replaced or scaled without disturbing the IT-side secondary loop plumbing. North America remains the dominant regional market across all reported scopes, reflecting hyperscale concentration in the US [S4].
Cost barrier and integration complexity: who CDU is NOT for
CDU capital cost runs higher than air-cooled equivalents because of specialized brazed-plate or coaxial heat exchangers, redundant pumps, leak-detection instrumentation, and BMS/EPMS integration, a barrier most acute for mid-sized operators and edge micro-data-centers below 2 MW IT load [S3]. For workloads that can stay below roughly 20-25 kW per rack, conventional air handling with hot-aisle/cold-aisle containment still has the lower total cost of ownership. CDU payback is driven by rack density crossing roughly 30-40 kW, where the chiller-side energy savings and the avoided CRAC capacity start outweighing the secondary-loop capex.
Side systems matter here: a properly specified distribution cabinet for the secondary loop, an FRL unit where compressed air is used for actuation, and a degassing unit on the facility loop to control dissolved oxygen and nitrogen are common supporting packages, though none of these alone justifies a CDU retrofit.
Standards and sourcing: what to pin on a spec

Data center CDU procurement in 2026 is dominated by Vertiv, Schneider Electric, Eaton, nVent, Nidec, Nortek DCC, CoolIT Systems, Coolcentric, Delta Electronics, DCX, Envicool, Kehua Data, and Chilldyne, per the IntelMarketR competitive landscape [S5]. Selection criteria buyers cite: precision coolant control (flow, pressure, temperature stability under step-load transients), modular scalability for staged rack fill, low-maintenance architecture (sealless pumps, accessible strainers), and digital monitoring with leakage and conductivity sensing [S5].
No single international standard governs CDU design end-to-end, so specifiers typically layer ASHRAE TC 9.9 thermal guidelines, manufacturer-specific secondary-loop coolant compatibility lists, and facility-level codes for pressure vessels and electrical safety on top of the OEM datasheet. Buyers should ask for a stated maximum rack power at the secondary-loop outlet, redundant pump configuration (N+1 or 2N), and a documented leak-detection response time before locking a vendor. For related reading on adjacent thermal and process instrumentation tracks, see dual-input temperature transmitter redundancy practice for 2026 and the asset/process/line twin decision matrix for 2026.
Trackable signals into 2027: a 2026 CDU market print in the USD 1.0B-11.4B band depending on scope, a data-center-only cumulative sales opportunity cited at USD 239.1 billion across 2026-2032 [S4], and a top-10 vendor concentration reported between 50% and 70% of 2025 revenue [S4]. Any of these three bands narrowing or widening in the next revision will be the cleanest read on whether liquid cooling is consolidating or fragmenting.