--- title: "Cooling & Resource Efficiency | Aguasto" description: "Aguasto’s evaluation framework for rack cooling, heat rejection, local and indirect water, electricity and quality-qualified compute. Design priorities, without operating environmental claims." canonical: "https://www.aguasto.com/infrastructure/cooling-resource-efficiency" last-updated: "2026-10-13" --- # Cooling & Resource Efficiency | Aguasto Canonical HTML: [Cooling & Resource Efficiency](https://www.aguasto.com/infrastructure/cooling-resource-efficiency) > Aguasto’s evaluation framework for rack cooling, heat rejection, local and indirect water, electricity and quality-qualified compute. Design priorities, without operating environmental claims. [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) Infrastructure / Cooling & resource efficiency ## Follow the heat. Define the resource boundary. An engineering framework for high-density AI infrastructure: rack cooling, final heat rejection, water, electricity and useful compute. These are Aguasto’s evaluation priorities; no operating efficiency result is claimed. Original Aguasto conceptual illustration. Not manufacturer geometry, an operating Aguasto asset, or a measured benchmark. The whole thermal path ### Liquid at the chip. A separate choice outside. Direct-to-chip cooling moves heat from selected components into a technology coolant circuit. A coolant distribution unit can transfer that heat across a controlled interface to the facility circuit. Other components may still require room air cooling. The facility must ultimately reject heat outside; dry, evaporative and hybrid arrangements have different climate, water, electricity, capacity and maintenance implications. Warm-water architectures can expand suitable operating options where the exact server, coolant, supply-temperature and ambient envelopes support them. They do not remove the need to validate the selected OEM equipment or establish a universal best design. Heat rejection / Engineering choices Follow the heat beyond the closed technology loop. Illustration: Two cooling circuits and dry heat rejection. A rack technology loop transfers heat across a CDU heat exchanger to a facility circuit. Final heat rejection and residual air cooling remain separate boundaries. No actual equipment settings or measured water use are shown. Original generic circuit study. The CDU separates fluids; it does not establish zero site water use. Colors show thermal paths, not measured flow, temperature or a project specification. Dry heat rejection Evaporative heat rejection Adiabatic / hybrid Dry heat rejection #### Reject heat to air; evaluate the temperature envelope. Water boundary Potentially reduces cooling-related local freshwater consumption compared with an evaporative configuration. Other site water uses and indirect electricity-related water remain separate. Design trade-offs Ambient conditions, approach temperatures, fan/pump electricity, capacity, space and capital can constrain the choice. Required rack supply conditions must still be met. Water and climate ### A closed loop is one boundary, not the whole site. #### Consumption and withdrawal Evaporation consumes water. Water withdrawn for use may include returned flows; makeup, blowdown, treatment and source conditions require their own accounting. Annual use cannot establish peak local demand. #### Reuse with conditions Reclaimed water may change freshwater demand where permitted and technically suitable. Quality, treatment, transport, availability and local water conditions need documented assessment. #### Climate-specific choices Dry systems can reduce cooling-related local freshwater use while changing electricity, approach temperatures, space, capital and thermal capacity. Hybrid water assistance needs explicit operating conditions. Aguasto has no verified water, PUE, WUE, carbon or token-efficiency results. Design intent is not a zero-water, water-neutral, net-zero or comparative environmental claim. Measurement framework ### Measure the facility. Then measure useful work. Site overhead, local water conditions and quality-qualified model output answer different questions. Comparisons require aligned measurement periods, disclosed equipment and operating boundaries, and reproducible workload conditions. **Metric** PUE **What to measure** Facility energy divided by IT energy over aligned measurement periods. **Required boundary** Identify the facility, meters, shared services and reporting period. PUE alone is not useful AI work or a water metric. **Metric** WUE **What to measure** Defined facility water use per unit of IT energy, with the applicable reporting boundary stated. **Required boundary** Identify units, period, water sources and what the metric includes. Keep withdrawal and consumption visible rather than interchanging them. **Metric** Local freshwater consumption **What to measure** Freshwater consumed within the stated on-site boundary. **Required boundary** Separate evaporation and other consumptive uses from withdrawn water that is returned; identify local water conditions. **Metric** Peak water withdrawal **What to measure** Water taken from a source over the specified peak interval. **Required boundary** Annual averages do not establish peak demand, utility availability or local environmental suitability. **Metric** Indirect water **What to measure** Electricity-related water effects outside the facility boundary. **Required boundary** State generation/location/time assumptions. A site cooling claim is not a complete system-wide water footprint. **Metric** Cooling electricity **What to measure** Measured energy for the included pumps, fans, chillers and related systems. **Required boundary** Identify which loops, residual air cooling and heat-rejection equipment are counted under the actual climate/load. **Metric** Useful compute per kWh **What to measure** Quality-qualified completed work or tokens per the defined energy denominator. **Required boundary** Specify checkpoint, request/context distribution, quality, latency, concurrency, errors and included equipment. No Aguasto result is available. **Metric** Facility energy per billable work **What to measure** Included facility energy per completed commercial workload unit. **Required boundary** Separate a metering definition from token billing rules. Idle energy, overhead and failed work need explicit treatment. **Metric** Embodied carbon & life cycle **What to measure** A disclosed lifecycle assessment of included construction, equipment, transport and replacement. **Required boundary** Declare inventory, service-life assumptions and exclusions. Reuse is not automatically lower impact or indefinitely feasible. For GPU work and tokens per kWh, pin the checkpoint, runtime, precision, input/output lengths, latency and quality criteria, concurrency, failed work and power metering. Device specifications and token billing are distinct from executed, facility-wide measurements. ### Continue the research [Illustrative electricity and completed-token economics](https://www.aguasto.com/tools) [Serving configuration and software trade-offs](https://www.aguasto.com/inference-stack) [Checkpoint and license boundaries](https://www.aguasto.com/models) Aguasto engineering priorities ### Designed for more resource-conscious AI infrastructure. - Evaluate appropriate reuse and avoidable construction waste against actual condition and safety. - Specify adaptable equipment interfaces and future replacement/transport limits. - Select qualified cooling and heat-rejection systems for the real climate and hardware envelope. - Consider local freshwater conditions alongside electricity-related and lifecycle impacts. - Improve useful GPU work through evaluated software, utilization and operating practices. - Measure operating outcomes and report verified results when facilities become operational. Evaluate construction, equipment, transport, replacement and end-of-life together. Suitable reuse and modular adaptation can avoid some work, but condition, safety, interface changes and service life determine actual outcomes. Qualified engineering and technical partners would validate functions beyond Aguasto’s established capabilities. Evidence & methodology ### Sources define the questions. Operating evidence must answer them. This framework draws on DOE’s facility energy and water guidance, LBNL’s data-center energy/water accounting and ASHRAE’s liquid-cooling system distinctions. Manufacturer material is attributed as vendor guidance; it does not verify an Aguasto result. [DOE FEMP: data centers and servers](https://www.energy.gov/cmei/femp/cooling-water-efficiency-opportunities-federal-data-centers) · [LBNL: 2024 energy and water study](https://escholarship.org/content/qt32d6m0d1/qt32d6m0d1.pdf) · [ASHRAE: liquid cooling guidance](https://handbook.ashrae.org/Handbooks/A23/IP/A23_Ch20/a23_ch20_ip.aspx) ### Continue the research [Why Modular and distributed resilience](https://www.aguasto.com/infrastructure/why-modular) [Rack density and cooling interfaces](https://www.aguasto.com/learn/liquid-cooling-gpu-rack-interfaces) [Air, liquid, dry and hybrid cooling guide](https://www.aguasto.com/learn/data-center-cooling-air-liquid-dry-hybrid) [Freshwater, withdrawal and consumption](https://www.aguasto.com/research/data-center-water-consumption-and-withdrawal) [PUE, WUE and useful compute measurement](https://www.aguasto.com/learn/measuring-pue-wue-and-ai-compute-efficiency) [Environmental trade-offs and equipment reuse](https://www.aguasto.com/research/environmental-tradeoffs-and-infrastructure-reuse) ## 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.