Joule Heist search

topics · updated 2026-09-25

Commercial Adsorption Chillers — Vendors, Specs & Deployments

confidence: medium volatility: hot verified: 2026-09-25stale

The buyable state of the art in adsorption chillers as of September 2026: who makes them (Fahrenheit/SorTech, InvenSor, Bry-Air, plus Mitsubishi AQSOA as an OEM adsorbent), their model specs and working pairs, the small ~$215 M market still running on silica gel and zeolite rather than MOFs, measured field deployments (CoolMUC-2, iDataCool), and the 2025–2026 shifts: a third-generation architecture being fielded at two industrial RE-WITCH sites, new entrants Cooll and Exora at the two ends the incumbents cannot reach, a marine pilot as Fahrenheit's only visible activity, and a data-center prize being taken by absorption and district-heat export rather than adsorption.

What you can actually buy today. The adsorption-chiller market is a small (~USD 215 M, 2025), slow-growing (~3.6 %/yr) niche, and every commercial machine still runs on silica gel or zeolite with a water (R718) refrigerant — not the lab MOFs. This article maps the current vendors, their model specs and driving temperatures, the real installations that prove the technology in the field, and what moved in 2025–2026. For the physics see Adsorption Cooling; for the quantitative envelope see Performance & Numbers. Vendor figures here come from datasheets, distributor pages and press coverage, so treat them as claims rather than measurements unless a field deployment says otherwise.

The market shape

The market is USD 215.4 M (2025) → USD 306.8 M (2035), CAGR 3.6 % — a genuine but slow-growing niche driven mostly by industrial waste-heat recovery (chemicals ~35 %, food ~25 %, oil & gas ~20 %), with the fastest growth in China (4.8 %) and India (4.5 %). Named incumbents: Hitachi (~24 % share), Johnson Controls/YORK, Fahrenheit, Bry-Air, Thermax, Kawasaki Thermal Engineering, Panasonic, Ebara, Carrier, Inoplex, Berg — several of which (Thermax, JCI/YORK, Carrier) are primarily absorption houses, so the list should be read as “thermally driven chillers” broadly. Crucially, the market analysis contains zero MOF mentions: what ships today is silica gel and zeolite. That gap is the subject of the MOF-commercialisation discussion in Performance & Numbers and the durability verdict in Limitations & Mitigations. The wider context from RE-WITCH: EU industry spends ~150 TWh/yr of electricity on process cooling against ~300 TWh/yr of available industrial waste heat, roughly 35 % of it below 200 °C — the addressable pool is enormous relative to the market that exists.

Vendors and product lines

VendorProduct lineWorking pairCapacityDriving tempCOP
Fahrenheit GmbH (ex-SorTech)eCoo S / 20 ST / 20X / 30 / 30X / 40XSilica gel (or zeolite) + water (R718); R290 on some heat-reject sides~10 kW (eCoo S) → ~40 kW per module50–95 °C~0.35–0.6 (eCoo 10 bench ~0.35–0.4), claimed ~90–95 % of Carnot across 60–90 °C
InvenSor GmbHLTC 10 e plus / LTC 30 e plus / HTC 18 plusDirect-coated zeolite + water10 / 10–35 / 18 kW45–100 °C (LTC nominal 72 °C in; HTC from 85 °C)0.52 nominal, up to 0.75; EER up to 33
Bry-Air / Pahwa (India)BryChillSilica gel (BRYSORB 200) + water11–35 kW; industrial to 1180 kW (335 TR)50–100 °C (regeneration from 60 °C)not published
Mitsubishi ChemicalAQSOA Z01/Z02/Z05 (adsorbent, not a chiller)Aluminophosphate zeoliteOEM materiallow regeneration (< 90 °C)—

Notes:

  • InvenSor distinguishes itself with direct-coated zeolite beds and the highest catalog COP (0.75 max; the LTC 10 e plus is rated 10 kW at 72 °C in, 10 °C chilled out, 31.3 °C recooling out, with chilled water 10–25 °C), plus a dedicated HTC line tuned for high ambient temperatures (> 40 °C) at COP ~0.5 from 85 °C heat — a productised answer to the hot-day derating problem. Caveat: the official datasheets are distributor-hosted image PDFs, so figures come from a distributor case study and search snippets.
  • Fahrenheit’s eCoo modules have no moving parts; the eCoo S brings the entry point to a refrigerator-sized 10 kW unit; chilled water 9–21 °C; integration via BACnet, Modbus, M-Bus, LON and OPC. The distributor page cites a 600 kW Fahrenheit reference at the Leibniz Supercomputing Centre — source-vs-source note: the LRZ’s own documentation describes the CoolMUC-2 installation as six eCoo 2.0 units delivering ~50 kW of cooling (below), so the 600 kW figure either refers to a different or aggregate LRZ system or is a distributor overstatement; treat as unverified. Fahrenheit’s own site blocks automated fetch.
  • Bry-Air bills BryChill as the first adsorption chiller launched in India: a four-chamber near-vacuum design (evaporator, condenser, two adsorption chambers), modular and stackable, with a ~25-year claimed lifespan.
  • AQSOA is the low-regeneration adsorbent that makes sub-90 °C machines possible; it is supplied into OEM chillers rather than sold as one, and it is the pair in the automotive research in Performance & Numbers.
  • Vehicle/mobile adsorption A/C is an active patent space. OxiCool Inc. (US 9,765,998 B2, granted Sep 2017, priority to a March 2013 provisional; 16 claims) covers a zeolite/water system with a combined evaporator/condenser, a two-bed layout alternating adsorption and recharge for continuous cooling, desorption from dedicated burners (natural gas, diesel, LP, jet fuel, solar, geothermal, battery) or truck-engine exhaust, and a “winterization configuration” in which the control system adsorbs the water into the zeolite’s crystalline structure when ambient approaches freezing so no ice can form — freeze protection without antifreeze. Control logic picks the winterization mode by how fast ambient is falling and can act on weather forecasts; the target is truck-sleeper cooling — the same niche as the AQSOA-Z02 automotive research (see Adsorbent Bed Engineering).

What’s coming: advanced-sorbent R&D

The clearest pre-commercial effort is the EU DYMAN project (EIC Pathfinder Challenges grant 101161930; three years; ten partners in Spain, Germany, Italy and Belgium including Sorption Technologies GmbH, the SorTech successor). It is developing adsorbents that work below 50 °C driving temperature — below even the EMM-8 regime — and 3D-printed adsorption heat exchangers that embed sorbent in a porous structure to cut internal thermal resistance and exploit two-phase flow, aiming to cool HPC data centers (a 200–500 TWh sector) from their own waste heat. As of mid-2026 it is at chiller-prototype stage, pre-commercial.

Field deployments (measured, not rated)

Real installations matter because rated COP and measured COP diverge, and because climate and driving temperature dominate performance.

  • CoolMUC-2 — Leibniz Supercomputing Centre (Munich). Six SorTech eCoo 2.0 silica-gel/water chillers driven by hot-water-cooled HPC racks (384 Lenovo NeXtScale nodes) — the only Top500 production system combined with adsorption refrigeration. Measured 2016: ~120 kW of waste heat at 45 °C → ~50 kW of cooling at 21 °C, with an electrical COP ~12 (cooling per kW of electricity), roughly 2–3× a mechanical chiller. The standout is the very low 45 °C driving temperature — most adsorption chillers want 65–90 °C. Operated nine years, decommissioned December 2024.
  • iDataCool — University of Regensburg + IBM. 216 hot-water-cooled nodes; ~70 °C coolant drove an InvenSor LTC 09. Measured chiller COP rose ~90 % going from 57 °C to 70 °C driving temperature — a concrete sensitivity curve — and ~25 % of input energy was reusable as cooling. Lesson learned: poor rack thermal insulation caused major heat loss; designing insulation early could roughly double the reuse fraction.
  • KACST adsorption desalination + cooling — Riyadh. Carried from the prior compile: a megawatt-scale demonstration producing up to 1 MW of cooling and up to 100 m³/day of desalinated water from one waste-heat cycle. Not traceable to this article’s raw set; the documented KAUST four-bed solar pilot is in Performance & Numbers.

2025–2026: what actually moved

An August 2026 sweep across vendor news, funding and project reports found the landscape shifting at the edges while the incumbents sat still:

  • The incumbents are quiet. No 2025–2026 product launches, funding or discontinuations surfaced for Fahrenheit, InvenSor, Nishiyodo/HIJC, Mayekawa or Mitsubishi AQSOA (absence of evidence, not proof of retreat). Fahrenheit’s only visible activity is the EU Engimmonia marine pilot: an optimised adsorption chiller on engine jacket-water heat aboard the ANEK ferry F/B Elyros, with CNR-ITAE support, still in monitored testing as of June 2026. The marine envelope — 70–90 °C jacket water → 5–7 °C chilled water, COP 0.5–0.7, up to 80 % less electricity than compression chillers, silica-gel/distilled-water, no moving parts — is driven by IMO EEXI/CII efficiency rules and by the fact that modern marine diesels reject more than half their fuel energy as heat. Date caution: the Fahrenheit “rack-integrated adsorption chiller” item (16 kW from 34 kW of hot water in a 19-inch rack) that keeps surfacing is from June 2019, not new.
  • New entrants target the two ends Fahrenheit-class machines can’t reach. Cooll (Enschede, NL; University of Twente spin-off) closed a ~€4.5 M seed in October 2025 to commercialise a gas- or hydrogen-fired adsorption-compressor heat pump — a sorbent bed cycled thermally to pressurise refrigerant, no mechanical compressor — as a wall-hung boiler replacement claiming 25–40 % lower gas use and CO₂ than a condensing boiler; manufacturing partners (ManGo, Pillen) and a Dutch installer network are in place, a ~100-unit Dutch deployment is targeted, and extreme-cold validation at GTI Energy (Chicago) signals North American intent. Heating-led, but a genuinely new commercial sorption architecture. Exora (heise, July 2026; CEO Alexis Van Wesemael) claims a MOF-based chiller driven by ~35 °C data-center coolant return — far below the ~50–60 °C floor of anything shipping — by positioning MOF step isotherms so desorption completes at 30–45 °C liquid-cooling return temperatures with no booster heat. It has a 3 kW prototype, targets a 100 kW production unit by 2028 and MW-range by 2030, claims ~10 % of a conventional chiller’s electricity, and pitches the same machine for upgrading 30–45 °C heat to district-heat grade. Vendor claims via journalism, framed around the 2025 Chemistry Nobel for MOFs; the precursor academic result is SJTU’s encapsulated-composite chiller at 50 °C drive (SCP 102 W·kg⁻¹, Energy 2024), which Exora would undercut by 15 °C. Unverified.
  • Third-generation architecture is real and being fielded. Gado, Palomba, Brancato, Yarce, Mittelbach and Frazzica (CNR-ITAE + SorptionEnergy; Renewable Energy 2026) define a generational taxonomy — 1st gen research prototypes; 2nd gen SorTech/Fahrenheit-class machines with separate evaporator/condenser vessels and large vacuum valves, commercially marginal at ~SCP 246 W·kg⁻¹ / COP 0.48 — and introduce a 3rd generation: paired adsorber modules with evaporation and condensation directly at the adsorber and liquid-phase refrigerant distribution, eliminating the large, expensive vacuum valves; validated with a dynamic model against a modular test rig and framed as what could “unlock full market potential.” Sorption Technologies GmbH (also in DYMAN) is deploying exactly this design (patent DE 10 2020 129 341 A1; targets thermal COP 0.65, electric COP > 20, > 17 kW·m⁻³, 75–90 °C drive) through the 27-partner Horizon Europe RE-WITCH project: a 200 kW adsorption chiller at a Polish brewery (~80 °C wastewater plus a 50–70 kW TVP Solar high-vacuum flat-plate field; wort cooling to 5 °C; ~64 MWh/yr of electricity saved) and an 80 kW containerised unit at a German pharma plant (waste heat from a 225 kW air compressor feeding a 500 kW central cooling plant; 7,500 h/yr; 110 MWh/yr saved, 29 % of cooling electricity, 50 t CO₂/yr). The other two RE-WITCH sites are absorption (Spanish dairy, Greek bio-refinery), and the project projects solar cooling could cover 25 % of EU industrial cooling by 2050. These are engineering projections, not yet measured — the sites to watch for data in late 2026.
  • The data-center prize is being taken by absorption, not adsorption. Johnson Controls productised waste-heat cooling for AI data centers in 2026: an April release claiming YORK absorption chillers on CHP waste heat cut chiller electricity > 90 %, plus liquid-cooling investments (Alloy Enterprises, Accelsius direct-to-chip — the higher coolant return temperatures that enable any sorption chiller downstream); then a July CHP + YORK absorption reference design converting ~57 % of wasted heat and, for a modelled 1 GW facility, reclaiming up to 97 MW of compute at PUE 1.23 with zero on-site water, −44 % cooling electricity and −43 % cooling CO₂ (all modelled, explicitly disclaimed). Meanwhile Nordic hyperscalers export waste heat to district heating instead of converting it to cold (Microsoft Espoo/Høje-Taastrup, Google Hamina — commissioned 2025–26; delivered heat 12–22 EUR/MWh). Adsorption’s data-center opening narrows to modular sub-100 kW units on sub-90 °C heat where no district network exists, and the sub-50 °C drive band absorption cannot touch — exactly where DYMAN and Exora aim.
  • Where sorption capital went in 2025: heat-driven heat pumps (Cooll; Stone Mountain $14.5 M Series B for gas absorption) and solid-state competitors (MagnoTherm $7.8 M, magnetocaloric) — not classic silica-gel/zeolite chiller vendors. The established adsorption-chiller market looks static.

See Also

Sources