--- title: "Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures | Aguasto" description: "Separate equipment heat capture from final heat rejection, then evaluate an architecture against climate, hardware, water and operating constraints." canonical: "https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid" last-updated: "2026-10-13" --- # Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures | Aguasto Canonical HTML: [Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid) > Separate equipment heat capture from final heat rejection, then evaluate an architecture against climate, hardware, water and operating constraints. [Aguasto](https://www.aguasto.com/) [learn](https://www.aguasto.com/learn) Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures [Resources](https://www.aguasto.com/resources) [GPU Library](https://www.aguasto.com/gpus) [Model Index](https://www.aguasto.com/models) [Inference Stack](https://www.aguasto.com/inference-stack) [Learn](https://www.aguasto.com/learn) [Insights](https://www.aguasto.com/insights) [Research](https://www.aguasto.com/research) [Engineering tools](https://www.aguasto.com/tools) Aguasto / learn ## Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures Separate equipment heat capture from final heat rejection, then evaluate an architecture against climate, hardware, water and operating constraints. Original Aguasto conceptual illustration. Not manufacturer geometry, an operating Aguasto asset, or a measured benchmark. By Aguasto Research · Published October 13, 2026 · Revision 1 Aguasto Research documents third-party hardware, models and software. References do not represent Aguasto inventory, evaluated deployments or available inference services. Illustration: Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures: original resource boundary study. Equipment heat capture and outdoor heat rejection are two choices. Three sink options require climate, temperature, water and electricity assessment; no option is universally best. Equipment heat capture and outdoor heat rejection are two choices. Three sink options require climate, temperature, water and electricity assessment; no option is universally best. Original Aguasto explanatory artwork; hypothetical boundaries, not an operating project specification. ### Make two decisions, not one Air, liquid, dry and hybrid are not four interchangeable alternatives. Air or liquid can describe how heat leaves computing equipment; dry, evaporative or hybrid can describe how that heat finally reaches the outdoor environment. ASHRAE distinguishes equipment interfaces from ultimate heat rejection. A liquid-cooled rack can therefore connect to a dry cooler, a cooling tower or a system using both. An air-cooled room can also depend on an evaporative plant elsewhere on the site. Draw the entire path before comparing proposals: compute components, equipment coolant or room air, distribution, heat exchangers, any refrigeration, and final heat rejection. Mark the water and electricity entering each stage. This prevents a procurement label from becoming an unsupported environmental claim. Aguasto intends to evaluate complete architectures; it has no operating cooling system or verified savings to present. [Data Centers and Telecommunication Facilities — ASHRAE Handbook 2023, chapter 20 (external source)](https://handbook.ashrae.org/Handbooks/A23/IP/A23_Ch20/a23_ch20_ip.aspx) ### Keep the remaining air load in the design Direct-to-chip cooling captures heat through cold plates attached to selected components. The liquid heat-capture fraction depends on the equipment configuration. Vertiv's technical guidance describes residual air loads and distinguishes a CDU that transfers heat to facility water from one that rejects heat into room air. A liquid-to-air CDU still needs adequate room cooling; bringing liquid to processors does not make the heat disappear. Request a configuration-specific heat balance rather than assuming a fixed percentage applies to every GPU system. Account for power supplies, memory, networking and other components that retain an air path. Evaluate normal operation, maintenance and a failed cooling component. A design that closes its liquid balance while overlooking residual room heat can have an attractive schematic and an incomplete operating envelope. [Liquid and Immersion Cooling Options for Data Centers — Vertiv (external source)](https://www.vertiv.com/en-ca/solutions/learn-about/liquid-cooling-options-for-data-centers/) ### Dry heat rejection is constrained by the temperature difference A dry heat exchanger uses ambient air as its sink. The process fluid must remain warmer than that sink by a useful heat-transfer difference. NREL's installation study explains this physical boundary; Vertiv's May 2026 analysis reinforces that coolant requirements and local ambient conditions jointly determine whether compressor-free operation is feasible. Warmer permitted coolant can expand the operating window, but it cannot make an unfavorable outdoor condition vanish. For a candidate site, examine hourly weather alongside the equipment's permitted supply temperature, exchanger approach and design load. Check the hottest required operating condition separately from a typical annual hour. Estimate fan and pump demand, required outdoor area and noise. If refrigeration or another assist closes the peak-condition gap, include it explicitly. Describe the architecture as dry in the hours when it actually operates dry. [Beyond the compressor-less hype: Matching data center cooling technology to climate reality — Vertiv (external source)](https://www.vertiv.com/en-us/insights/articles/educational-articles/beyond-the-compressor-less-hype-matching-data-center-cooling-technology-to-climate-reality/) · [Thermosyphon Cooler Hybrid System: Modeling and Installation — NREL/CP-2C00-66690, February 2017 (external source)](https://docs.nlr.gov/docs/fy17osti/66690.pdf) ### Distinguish evaporative and adiabatic assistance Evaporative systems use water to improve heat rejection. A cooling tower loses water through evaporation, drift and blowdown; leaks add avoidable losses. FEMP explains that controlling dissolved solids constrains how far blowdown can be reduced. Adiabatic assistance can improve the inlet-air condition of a dry-cooling arrangement, but its wet operating periods must still appear in the water balance. Compare the whole operating schedule: fully dry hours, assisted hours and any mechanical-cooling hours. Specify water quality, makeup capacity, treatment, discharge and maintenance requirements. Reclaimed water may change the freshwater source without changing the amount evaporated. The decision is whether the additional thermal capability justifies the local water, operational and financial requirements under the selected site's actual constraints. [Beyond the compressor-less hype: Matching data center cooling technology to climate reality — Vertiv (external source)](https://www.vertiv.com/en-us/insights/articles/educational-articles/beyond-the-compressor-less-hype-matching-data-center-cooling-technology-to-climate-reality/) · [Best Management Practice #10: Cooling Tower Management — DOE FEMP (external source)](https://www.energy.gov/cmei/femp/best-management-practice-10-cooling-tower-management) ### A hybrid system needs a tested control sequence The NREL thermosyphon work provides a useful evidence sequence: a February 2017 modeling and installation paper, followed by a September 2018 report describing 24 months of results. The system combined a dry device with an open cooling tower and coordinated their use. It demonstrates a particular implemented architecture and measured period; it does not establish a universal result for every climate, workload or modern GPU installation. Treat hybrid control as an engineering deliverable. Define which conditions change the mode, how equipment stages, what happens when a sensor fails, and which mode remains available during maintenance. Ask whether water restrictions coincide with extreme heat and whether the fallback still carries the required load. Commission transitions as well as steady states. A collection of capable components becomes a reliable cooling system only when the interfaces and control behavior are established. [Thermosyphon Cooler Hybrid System: Modeling and Installation — NREL/CP-2C00-66690, February 2017 (external source)](https://docs.nlr.gov/docs/fy17osti/66690.pdf) · [Thermosyphon Cooler Hybrid System: Results from 24 Months and Impact on WUE — NREL/TP-2C00-72196, September 2018 (external source)](https://www.nlr.gov/docs/fy18osti/72196.pdf) ### Start from the exact equipment envelope ASHRAE's liquid-cooling material discusses facility temperatures, equipment requirements and the implications of changing those interfaces. Its classes and examples are useful engineering context, not permission to run an unspecified system at an assumed temperature. Equipment limits, coolant quality, flow and pressure requirements need to come from the selected configuration's documentation and approved design. Record supply and return temperatures at both sides of each heat exchanger. Include fluid composition, material compatibility, dew-point protection where relevant, filtration and service procedures. Avoid substituting a chip junction temperature for a coolant supply temperature. Keep unknown configuration details open until the manufacturer and qualified engineering team confirm them. Selecting a cooling architecture before establishing these requirements can conceal the most important constraint. [Data Centers and Telecommunication Facilities — ASHRAE Handbook 2023, chapter 20 (external source)](https://handbook.ashrae.org/Handbooks/A23/IP/A23_Ch20/a23_ch20_ip.aspx) · [Emergence and Expansion of Liquid Cooling in Mainstream Data Centers — ASHRAE, 2021 (external source)](https://www.ashrae.org/file%20library/technical%20resources/bookstore/emergence-and-expansion-of-liquid-cooling-in-mainstream-data-centers_wp.pdf) ### Select the architecture for a phase and a place The ASHRAE handbook cautions that an installation sized for its final load may operate differently at a small initial load. For modular deployment, evaluate the first phase, intermediate additions and the intended full configuration. Standardized module interfaces can simplify repetition, while outdoor heat rejection, water availability and climate remain site-specific decisions. A useful selection record contains the equipment envelope, design and partial-load cases, annual cooling electricity, site-water demand, peak makeup requirement, maintenance coverage, redundancy assumptions and installed cost. Use the same workload and reliability requirement for each alternative. State which inputs are modeled and which are measured. Revisit the record when the rack configuration or expansion plan changes; the preferred architecture is conditional on those inputs, not a permanent ranking of cooling technologies. [Data Centers and Telecommunication Facilities — ASHRAE Handbook 2023, chapter 20 (external source)](https://handbook.ashrae.org/Handbooks/A23/IP/A23_Ch20/a23_ch20_ip.aspx) · [Beyond the compressor-less hype: Matching data center cooling technology to climate reality — Vertiv (external source)](https://www.vertiv.com/en-us/insights/articles/educational-articles/beyond-the-compressor-less-hype-matching-data-center-cooling-technology-to-climate-reality/) ### Scope & limitations - This is an architecture-selection framework, not a detailed mechanical design or equipment recommendation. - Historical laboratory cases and vendor analyses retain their stated climate, configuration and measurement boundaries. - No Aguasto cooling configuration, water outcome, PUE, deployment timeline or hardware compatibility has been tested or claimed. ### Continue the research [Liquid Cooling for High-Density GPU Systems: Follow the Interfaces](https://www.aguasto.com/learn/liquid-cooling-gpu-rack-interfaces) [How Modular AI Data Centers Are Built and Commissioned](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning) [Why Electricity Availability Shapes AI Infrastructure Deployment](https://www.aguasto.com/research/electricity-availability-and-time-to-power) [The Aguasto modular thesis](https://www.aguasto.com/infrastructure/why-modular) [Potential OEM and strategic collaboration](https://www.aguasto.com/partners) ### Sources & verification The register preserves each source's publication and verification date. Manufacturer claims and documentation are attributed; Aguasto has not run the cited products in production. - [Data Centers and Telecommunication Facilities — ASHRAE Handbook 2023, chapter 20 ↗](https://handbook.ashrae.org/Handbooks/A23/IP/A23_Ch20/a23_ch20_ip.aspx) Air and liquid equipment interfaces; Thermal envelopes and minimum-load design; Dry, wet and hybrid heat rejection. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Liquid and Immersion Cooling Options for Data Centers — Vertiv ↗](https://www.vertiv.com/en-ca/solutions/learn-about/liquid-cooling-options-for-data-centers/) Residual air loads depend on heat capture; Liquid-to-air and liquid-to-liquid CDUs; Capacity balancing and heat rejection. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Beyond the compressor-less hype: Matching data center cooling technology to climate reality — Vertiv ↗](https://www.vertiv.com/en-us/insights/articles/educational-articles/beyond-the-compressor-less-hype-matching-data-center-cooling-technology-to-climate-reality/) Climate and coolant conditions constrain compressor-free cooling; Dry, adiabatic and mechanical operating modes; Site water and energy trade-offs. Publication: 2026-05-14. Verified: 2026-10-13. - [Thermosyphon Cooler Hybrid System: Modeling and Installation — NREL/CP-2C00-66690, February 2017 ↗](https://docs.nlr.gov/docs/fy17osti/66690.pdf) Dry device and open tower integration; Hourly modeled energy/water behavior; Installation study distinguished from later measured results. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Best Management Practice #10: Cooling Tower Management — DOE FEMP ↗](https://www.energy.gov/cmei/femp/best-management-practice-10-cooling-tower-management) Evaporation, drift, blowdown and leaks; Water chemistry limits concentration cycles; Makeup and blowdown metering. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Thermosyphon Cooler Hybrid System: Results from 24 Months and Impact on WUE — NREL/TP-2C00-72196, September 2018 ↗](https://www.nlr.gov/docs/fy18osti/72196.pdf) Measured results from the ESIF case; PUE, ERE and WUE tracked together; Site-specific hybrid operation. Publication: date not stated in the reviewed source. Verified: 2026-10-13. - [Emergence and Expansion of Liquid Cooling in Mainstream Data Centers — ASHRAE, 2021 ↗](https://www.ashrae.org/file%20library/technical%20resources/bookstore/emergence-and-expansion-of-liquid-cooling-in-mainstream-data-centers_wp.pdf) Facility and equipment temperature boundaries; Liquid-to-air interim architecture; Wetted-material compatibility and maintenance. Publication: date not stated in the reviewed source. Verified: 2026-10-13. ### In this publication [Make two decisions, not one](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-1) [Keep the remaining air load in the design](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-2) [Dry heat rejection is constrained by the temperature difference](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-3) [Distinguish evaporative and adiabatic assistance](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-4) [A hybrid system needs a tested control sequence](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-5) [Start from the exact equipment envelope](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-6) [Select the architecture for a phase and a place](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#section-7) [More learn ↗](https://www.aguasto.com/learn) [Source register ↓](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid#sources) ## Company navigation - [Home](https://www.aguasto.com/) - [Infrastructure](https://www.aguasto.com/infrastructure) - [GPU Cloud](https://www.aguasto.com/gpu-cloud) - [AI Inference](https://www.aguasto.com/ai-inference) - [Development & Investment](https://www.aguasto.com/development) - [Company](https://www.aguasto.com/company) - [Contact](https://www.aguasto.com/contact) - [Privacy](https://www.aguasto.com/privacy) - [Resources](https://www.aguasto.com/resources) - [GPU Library](https://www.aguasto.com/gpus) - [Model Index](https://www.aguasto.com/models) - [Inference Stack](https://www.aguasto.com/inference-stack) - [Learn](https://www.aguasto.com/learn) - [Insights](https://www.aguasto.com/insights) - [Research](https://www.aguasto.com/research) - [Industry cases](https://www.aguasto.com/research/cases) - [Why Modular](https://www.aguasto.com/infrastructure/why-modular) - [Cooling & Resource Efficiency](https://www.aguasto.com/infrastructure/cooling-resource-efficiency) - [Partners](https://www.aguasto.com/partners) - [GPU comparison](https://www.aguasto.com/gpus/compare) - [Model comparison](https://www.aguasto.com/models/compare) - [Engineering tools](https://www.aguasto.com/tools) - [Development friction](https://www.aguasto.com/research/hyperscale-development-friction) - [NVIDIA H100 SXM](https://www.aguasto.com/gpus/h100) - [NVIDIA H200 SXM](https://www.aguasto.com/gpus/h200) - [NVIDIA B200 SXM](https://www.aguasto.com/gpus/b200) - [NVIDIA B300](https://www.aguasto.com/gpus/b300) - [NVIDIA GB200 NVL72](https://www.aguasto.com/gpus/gb200-nvl72) - [NVIDIA GB300 NVL72](https://www.aguasto.com/gpus/gb300-nvl72) - [NVIDIA Vera Rubin NVL72](https://www.aguasto.com/gpus/vera-rubin-nvl72) - [AMD Instinct MI300X](https://www.aguasto.com/gpus/mi300x) - [AMD Instinct MI325X](https://www.aguasto.com/gpus/mi325x) - [AMD Instinct MI355X](https://www.aguasto.com/gpus/mi355x) - [AMD Instinct MI455X](https://www.aguasto.com/gpus/mi455x) - [AMD Helios](https://www.aguasto.com/gpus/helios) - [GLM-4.5](https://www.aguasto.com/models/glm-4-5) - [Qwen3-235B-A22B](https://www.aguasto.com/models/qwen3-235b-a22b) - [DeepSeek-R1](https://www.aguasto.com/models/deepseek-r1) - [Llama 3.3 70B Instruct](https://www.aguasto.com/models/llama-3-3-70b-instruct) - [Gemma 3 27B IT](https://www.aguasto.com/models/gemma-3-27b-it) - [gpt-oss-120b](https://www.aguasto.com/models/gpt-oss-120b) - [Kimi-K2-Instruct](https://www.aguasto.com/models/kimi-k2-instruct) - [GLM-5.3](https://www.aguasto.com/models/glm-5-3) - [Qwen3.8-27B](https://www.aguasto.com/models/qwen3-8-27b) - [DeepSeek-V4.1-Flash](https://www.aguasto.com/models/deepseek-v4-1-flash) - [Llama 4 Scout 17B-16E Instruct](https://www.aguasto.com/models/llama-4-scout-17b-16e-instruct) - [Gemma 4 31B IT](https://www.aguasto.com/models/gemma-4-31b-it) - [Kimi K3](https://www.aguasto.com/models/kimi-k3) - [vLLM](https://www.aguasto.com/inference-stack/vllm) - [SGLang](https://www.aguasto.com/inference-stack/sglang) - [TensorRT-LLM](https://www.aguasto.com/inference-stack/tensorrt-llm) - [NVIDIA Dynamo](https://www.aguasto.com/inference-stack/dynamo) - [NVIDIA Triton Inference Server](https://www.aguasto.com/inference-stack/triton) - [AMD ROCm](https://www.aguasto.com/inference-stack/rocm) - [Ray Serve](https://www.aguasto.com/inference-stack/ray-serve) - [Kubernetes GPU orchestration — NVIDIA device plugin](https://www.aguasto.com/inference-stack/kubernetes) - [Model weights are only the beginning of the memory budget](https://www.aguasto.com/learn/model-weights-and-runtime-memory) - [Why tokens per second is not enough to benchmark inference](https://www.aguasto.com/research/inference-benchmark-methodology) - [Prefill, decode and the cost of separating them](https://www.aguasto.com/learn/prefill-decode-and-disaggregated-serving) - [Choosing between vLLM, SGLang and TensorRT-LLM](https://www.aguasto.com/learn/choosing-an-inference-serving-framework) - [From Power to Inference—Building a Distributed AI Compute Network](https://www.aguasto.com/learn/from-power-to-inference) - [Dedicated compute and shared inference: compare the whole workload](https://www.aguasto.com/insights/dedicated-and-shared-inference-economics) - [Distributed inference reliability begins with the failure boundary](https://www.aguasto.com/learn/distributed-inference-reliability) - [Core Scientific: reading contracted, energized and billed capacity](https://www.aguasto.com/research/cases/core-scientific-capacity-disclosures) - [Why Modular AI Infrastructure Matters Now](https://www.aguasto.com/research/why-modular-ai-infrastructure-matters-now) - [What Is a Modular AI Data Center?](https://www.aguasto.com/learn/what-is-a-modular-ai-data-center) - [How Modular AI Data Centers Are Built and Commissioned](https://www.aguasto.com/learn/modular-data-center-factory-to-commissioning) - [Liquid Cooling for High-Density GPU Systems: Follow the Interfaces](https://www.aguasto.com/learn/liquid-cooling-gpu-rack-interfaces) - [Why Electricity Availability Shapes AI Infrastructure Deployment](https://www.aguasto.com/research/electricity-availability-and-time-to-power) - [What Is KV Cache? Capacity, Reuse and Isolation](https://www.aguasto.com/learn/kv-cache-capacity-and-policy) - [Continuous Batching: Throughput Without Hiding Latency](https://www.aguasto.com/learn/continuous-batching-and-latency) - [Applied Digital: Reading the Polaris Forge 1 Delivery Milestones](https://www.aguasto.com/research/cases/applied-digital-polaris-forge-capacity-milestones) - [Crusoe Abilene: Separate Delivered Infrastructure from Expansion Plans](https://www.aguasto.com/research/cases/crusoe-abilene-delivery-and-expansion) - [Data Center Cooling: Choosing Air, Liquid, Dry and Hybrid Architectures](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid) - [Can Data Centers Reduce Water Consumption? Define the Water Boundary First](https://www.aguasto.com/research/data-center-water-consumption-and-withdrawal) - [Measuring PUE, WUE and AI Compute Efficiency at the Right Boundary](https://www.aguasto.com/learn/measuring-pue-wue-and-ai-compute-efficiency) - [The Environmental Trade-offs of High-Density AI Infrastructure](https://www.aguasto.com/research/environmental-tradeoffs-and-infrastructure-reuse) - [Official LinkedIn](https://www.linkedin.com/company/aguasto) This is a generated representation of the public HTML page. Diagrams are conceptual. GPU and inference services are planned and are not currently available to book or call. The contact page describes the available inquiry flow.