The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →No—not in any universal sense. Liquid cooling is gaining ground as AI and high-performance computing push rack heat loads higher, but it is not a guaranteed win for energy, water, cost, or reliability. The result depends on how heat is captured at the servers, how the facility rejects it, and which impacts are counted. A design can reduce water used onsite yet shift some water demand to electricity generation; another can use more local water while requiring less energy. Those are real trade-offs, not proof that every gain must be paid for with an equal loss.
Why “zero-sum” is the wrong default
The phrase bundles together three separate questions: can liquid remove heat from dense computing equipment, does the complete facility use less energy, and does a particular design reduce total water or other resource impacts? A positive answer to the first does not settle the other two.
Liquid can carry heat away from chips more effectively than air in suitable designs, potentially reducing fan and cooling loads or enabling higher rack densities. But that heat still has to leave the building. The facility’s heat-rejection equipment, local climate, power supply, water availability, and the comparison baseline all affect the outcome. The International Energy Agency’s 4E Energy Efficient End-use Equipment (EDNA) report discusses potential energy savings while emphasizing that results and metrics depend on system boundaries. IEA 4E EDNA, “Liquid Cooling in Data Centres” (22 June 2026)
What “liquid cooling” means in a data center
Liquid cooling is a family of ways to capture heat from IT equipment. It is not a single system, and the IT-side method does not dictate how the building ultimately rejects that heat.
#1 Best Overall
| Approach | How it captures heat | Key distinction |
|---|---|---|
| Rear-door heat exchanger | A coil at the rack captures heat from server exhaust air. | Air still carries heat out of the servers; the rack-mounted coil transfers it to liquid. |
| Cold plates (direct-to-chip) | Liquid channels in plates replace conventional heat sinks on selected chips. | Typically uses a technology loop connected through a coolant distribution unit (CDU). |
| Immersion | Electronics sit in nonconductive dielectric fluid. | Single-phase systems circulate fluid; two-phase systems use boiling and condensation in a closed cycle. |
These categories and descriptions follow the U.S. Department of Energy’s data-center design guidance. Actual installations can combine approaches or leave some components air cooled. U.S. DOE FEMP, “Best Practices Guide for Energy-Efficient Data Center Design” (July 2024)
The CDU and the facility loop
In many direct-liquid designs, a recirculating technology loop carries heat from the servers to a CDU. The CDU transfers that heat to a facility loop and manages coolant flow. From there, a building may use chillers and cooling towers, dry coolers, or a hybrid arrangement to release heat outdoors. So “liquid cooled” does not automatically mean water-free, nor does a closed loop at the rack guarantee a closed loop for the whole facility. DOE FEMP’s design guide and DOE FEMP’s cooling-water guidance distinguish IT cooling from facility heat rejection.
Adoption is rising, but forecasts are not installed capacity
Available evidence points to growing interest rather than universal conversion. The figures below describe different kinds of evidence, so they should not be read as one continuous measure of adoption.
| Source and evidence type | Reported figure | What it does—and does not—show |
|---|---|---|
| IEA 4E EDNA, publication page, 2026 | Potential savings of about 8% in server energy, 30–40% at facility level, and 10–21% overall. | These are study potentials at different boundaries, not guaranteed savings for a particular operator. |
| S&P Global, enterprise decision-maker survey, 2026 | 21% planned to shift to liquid cooling over the next year; another 25% planned to switch in two to four years. The next-year share was 13% in its 2024 survey. | Respondents’ plans indicate intent, not completed deployments or installed-capacity share. |
| TrendForce, market forecast, 2025 | Forecast liquid-cooling penetration in AI data centers of 14% in 2024 and 33% in 2025. | These are market-analysis projections, not a verified census of operating systems. |
Sources: IEA 4E EDNA, S&P Global, and TrendForce. TrendForce cites 130–140 kW rack thermal design power (TDP) for NVIDIA GB200/GB300 NVL72 systems as an example of the density pressure behind early liquid-to-air deployments; that is a cited system example, not a universal rack load. The IEA reported data-center electricity demand grew 17% in 2025, while also identifying grid-connection, planning, and supply-chain bottlenecks. That wider demand context is not a measure of electricity used by cooling alone. IEA, “Data centre electricity use surged in 2025…” (2026)
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Does liquid cooling save energy?
It can, but the comparison has to cover the whole system and use a like-for-like workload and operating period. Chip-level heat capture may reduce server fans or facility cooling demand, while pumps, chillers, and heat-rejection equipment also consume energy. A reported saving at the server boundary cannot be treated as the same percentage saving for the whole building.
Two common metrics answer different questions. Power usage effectiveness (PUE) is annual total facility energy divided by annual IT equipment energy. Water usage effectiveness (WUE) is annual site water use divided by IT energy. PUE does not measure water or total environmental impact, and WUE is site-based: it does not by itself include water used to generate the facility’s electricity. Report either metric with its boundary and measurement period. The IEA 4E EDNA report also cautions that PUE may not fully express liquid cooling’s efficiency benefits. DOE FEMP’s design guide; IEA 4E EDNA
Rank #3
Does liquid cooling save water?
Not necessarily. A dry or closed-loop heat-rejection design can greatly reduce onsite cooling-water use, but it may require more electricity than an evaporative approach. Evaporative heat rejection uses local water, while potentially lowering energy demand. Whether the overall water impact improves depends in part on the grid’s generation mix and on the accounting boundary.
The California Energy Commission’s CalNEXT report estimates onsite data-center water consumption at 66 billion liters in 2023 and projects 150–280 billion liters by 2028. It also estimates indirect water consumption from electricity generation alone at nearly 800 billion liters in 2023, compared with 66 billion liters used directly for cooling. These are report-level estimates and projections, not measurements of the effect of liquid cooling; the direct and indirect figures cover different parts of the water footprint. The report also identifies supply-chain impacts, including component manufacturing, as outside site-only metrics. CalNEXT, “Datacenter Liquid Cooling Market Characterization: Final Report” (December 2025)
For a meaningful water comparison, separate onsite withdrawals and consumption from water associated with electricity generation and equipment supply chains. Pairing a site WUE figure with information about power sourcing gives a fuller picture than treating that metric as a total-water score.
Rank #4
What can prevent a good design from being a good project?
The best heat-capture method on paper may still be difficult to deploy or operate at a particular site. The IEA 4E EDNA report identifies standardization, initial cost, and long-term reliability concerns among reasons uptake remains low, and notes the need for retrofit solutions in existing multistorey data centers. Retrofitting can involve piping, space, power, service access, and changes to maintenance practices; the required work depends on the building and chosen architecture.
Facility constraints matter too. Power availability remains a leading data-center site-selection consideration in CBRE’s North America H1 2025 report. Local grid capacity, connection timelines, electricity prices, climate, water availability, and permitting can change both project feasibility and the trade-offs between evaporative and dry heat rejection. CBRE, “North America Data Center Trends H1 2025”
Liquid cooling is not the only lever. DOE guidance for existing cooling-tower systems includes air-side and water-side economizing and increasing cooling-tower cycles of concentration. Under suitable conditions, a water-side economizer can bypass chiller compressor load. These measures can complement a liquid-cooling deployment or improve a facility that retains air cooling. DOE FEMP, “Cooling Water Efficiency Opportunities for Federal Data Centers”
Best Value
- Data Center Coolant
- 25% Inhibited Propylene Glycol
- JeffCool ISF 25
- High thermal conductivity
How to judge a liquid-cooling proposal for a specific site
Ask the vendor and facility team to state assumptions and boundaries before comparing options. A useful comparison covers:
- Workload and rack density: What heat load must be removed, and is air cooling insufficient for the planned deployment?
- Heat capture: Is the design a rear-door exchanger, cold plate, immersion, or hybrid, and which components does it cool?
- Heat rejection: Does the building ultimately rely on cooling towers, chillers, dry coolers, or a hybrid system?
- Energy: Compare IT, fan, pump, chiller, and whole-facility energy against the same workload and a stated baseline.
- Water: Separate onsite use from electricity-generation and supply-chain water impacts; account for local scarcity.
- Retrofit fit and operations: Identify building changes, redundancy, leak management, maintenance procedures, service access, and downtime implications.
- Economics and schedule: Compare capital and operating costs, equipment lead times, and expected service life using the project’s assumptions.
- Location: Check climate, water availability, grid capacity, tariffs, connection timing, and permitting.
- Evidence quality: Label each result as measured, modeled, forecast, or based on survey intentions; state metric boundaries and operating periods.
No single controlled cross-technology comparison in the cited sources establishes a universal winner for every workload and climate. The sound conclusion is conditional: liquid cooling can solve heat-density problems and may improve energy performance, but its total resource effect depends on the full site design and what the comparison counts.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
