Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsLiquid cooling does not automatically stop when utility power fails, and it does not automatically keep working either. Cooling continues only if the pumps, controls and heat-rejection equipment in the complete cooling chain remain powered or transfer to a redundant path. If circulation or heat rejection is lost while servers continue running, their temperature can rise very quickly. Depending on the installation, backup power may preserve cooling, switch to another pump or coolant distribution unit (CDU), or provide enough continuity for an orderly reduction of IT load and shutdown.
The deciding factor is the site’s architecture—not the fact that it uses liquid. A UPS connected to one CDU controller is not proof that the facility pumps, chillers, cooling towers or air handlers will remain available.
As an Amazon Associate I earn from qualifying purchases.
The short answer: outage behavior is conditional
A liquid-cooled rack normally depends on more than the cold plates attached to its servers. A CDU circulates coolant through the technology loop, transfers heat across a heat exchanger and monitors flow, pressure and temperature. Facility-side equipment then carries that heat away. Every powered component in that path matters.
- Cooling remains active: UPS- or generator-backed pumps and controls continue circulating coolant, while the facility heat-rejection path remains available.
- Cooling fails over: a standby pump, CDU or power feed starts automatically when the primary component or supply is lost.
- Cooling is interrupted: a required pump, controller or heat-rejection device loses power without a usable alternate path.
- IT shuts down in a controlled way: if cooling cannot be sustained, site controls reduce load or turn off servers before thermal limits are reached.
These are possible sequences, not a universal progression. Automatic transfer switches, UPS capacity, generator topology, redundancy and control programming determine which one occurs.
#1 Best Overall
Which parts of a liquid-cooling system need power?
The CDU and technology loop
The CDU’s pump moves coolant through server cold plates or another direct-to-chip interface. Its controller reads temperature, pressure and flow sensors and operates valves, alarms and interlocks. Losing power to the pump stops or weakens active circulation; losing controller power can prevent a healthy redundant pump from starting even if that pump has electricity.
The facility heat-rejection loop
In a liquid-to-liquid design, a facility-water loop receives heat from the CDU’s heat exchanger. Facility pumps, chillers, dry coolers or cooling towers may be required to keep that water within its operating range. In a liquid-to-air arrangement, the CDU depends on the data center’s air-conditioning system to reject heat. A powered CDU with no available heat sink cannot provide indefinite cooling.
Power transfer and monitoring
An automatic transfer switch (ATS) can move a load from utility power to a generator or alternate source. UPS equipment can bridge an interruption and support controls or pumps, but runtime and connected loads vary by site. Alarm gateways, building-management systems and remote telemetry also need power if operators are to see flow, temperature and power status during the event.
Four outage outcomes operators may see
1. Cooling stays on through the event
A UPS or backed-up distribution system may keep CDU pumps and controls running while generators support longer-duration loads. Trane’s CDU-1MW documentation, for example, describes an ATS, UPS-backed controller continuity and two concurrently maintainable pumps. Those are features of that design, not a promise that every CDU has the same protection. If the facility-side cooling plant is also backed up, liquid circulation can continue while utility power is unavailable.
2. A redundant component takes over
Some installations use dual pumps, multiple CDUs or separate power feeds. Control logic detects a failed pump or unit and starts the standby path. Vertiv describes automated failover for redundant components and units, while ASHRAE guidance recommends fault-tolerant CDU design and backed-up CDU power. A failover is useful only when the standby equipment, valves, controls and downstream heat-rejection path are all ready and correctly configured.
3. Flow or heat rejection is interrupted
If a required pump, controller, chiller, tower, dry cooler or air handler loses power without an alternate path, active cooling can be reduced or stop. Servers that remain powered continue producing heat. ASHRAE’s liquid-cooling guidance warns that equipment temperature can rise very quickly after a CDU failure. The available material does not establish a universal time-to-overheat figure: the result depends on server heat output, coolant volume, flow rate, thermal limits, ambient conditions and protective controls.
4. IT load is reduced or shut down
When the cooling chain cannot be maintained, operators may shed workloads, power down selected racks or perform a wider graceful shutdown. ASHRAE recommends backing CDU power with a UPS so equipment can continue running or shut down cleanly during an outage. The trigger should come from the site’s operating procedure and measured conditions, such as loss of flow, rising supply temperature or a facility alarm—not from an assumed countdown.
Recommended Free Tools
Why there is no universal “minutes until overheating” answer
Liquid cooling removes heat efficiently, but efficiency does not create a standard emergency runtime. A high-density accelerator rack can add heat faster than a lightly loaded server rack. A loop with substantial coolant volume may have more thermal inertia than a compact loop. The CDU may continue pumping while the facility chiller is offline, or it may stop immediately if its controller loses power. Sensors may command a load reduction before hardware reaches a critical temperature, while a failed sensor or control link can delay that response.
Rank #3
For those reasons, a data center should establish limits from its own commissioning tests, equipment specifications and control sequences. Do not publish or rely on a generic number of seconds or minutes unless it is tied to a named system, test condition and operating state.
Architecture determines what the outage looks like
Liquid-to-liquid cooling
Here, the CDU transfers heat from the technology loop to facility water. The IT loop may be intact while the facility loop is unavailable, but servers still cannot run indefinitely because heat has nowhere to go. Resilience planning must therefore include facility pumps, heat exchangers, chillers and outdoor heat-rejection equipment, not just the rack-side CDU.
Liquid-to-air cooling
A liquid-to-air CDU rejects heat through the data center’s air-conditioning system. The outage question shifts to whether that air system, its fans and its controls have backup power. A generator-backed CDU paired with an unpowered air-conditioning plant can still lose effective heat rejection.
Free tools Windows power users keep installed
One-click scans. No signup required.
Loop arrangement and CDU location
CDUs may be installed in a row, at the rack or in a central plant. Loop length, isolation valves and the number of connected racks affect how a failure propagates. ASHRAE notes that CDU implementations and functionality vary. Vendor-specific valve positions or backup-CDU switchover behavior must be treated as design details, not universal liquid-cooling rules. Chilldyne’s published FAQ, for example, describes valve and switchover behavior for its own system only.
Rank #4
Backup power and redundancy: what each feature actually protects
| Feature | What it can preserve | What it does not prove |
|---|---|---|
| UPS-backed CDU controller | Control logic, sensor readings and orderly transfer during a power interruption | That pumps or facility heat rejection have UPS or generator power |
| UPS-backed CDU pump | Technology-loop circulation for the UPS-supported period | That the heat exchanger has an operating facility-side heat sink |
| Automatic transfer switch | Transfer to a configured alternate source | That the alternate source has enough capacity or starts successfully |
| Dual or concurrently maintainable pumps | Continued circulation after a single-pump fault or planned maintenance | Protection from a shared control panel, pipe, valve or power-feed failure |
| Redundant CDUs | Failover to another cooling unit when controls and isolation are functional | Automatic failover unless the sequence has been tested and enabled |
| Generator-backed facility plant | Chillers, towers, dry coolers, pumps or air handlers needed for heat rejection | That rack-level controls and local distribution are also backed up |
Trane and Eaton document examples of ATS, UPS, pump and heat-transfer features. Use those descriptions to identify questions for a design review, not to infer that every manufacturer’s equipment behaves identically.
What to verify before an outage
Document the complete electrical and thermal dependency chain for each liquid-cooled zone. The following checks should be part of commissioning and periodic drills.
- Trace every load. Record which UPS, generator bus or alternate feeder supplies each CDU controller, pump, sensor panel and local valve actuator.
- Trace facility heat rejection. Identify the facility pumps, chillers, cooling towers, dry coolers, heat exchangers and air handlers required for the zone, then verify their backup source.
- Confirm transfer behavior. Determine whether the ATS transfers automatically, which alarms appear, and whether the CDU remains in a safe operating mode during the transition.
- Test redundancy. Simulate a pump or CDU loss under an approved procedure. Confirm that the standby unit starts, valves move correctly and flow reaches the intended racks.
- Check alarm paths. Verify high supply temperature, low flow, pressure, pump fault and power-loss alarms locally and in remote telemetry. Confirm that monitoring remains available on backup power.
- Set operating thresholds. Define when to reduce IT load, isolate a rack or begin a graceful shutdown. Base the thresholds on manufacturer limits and measured site behavior.
- Write ownership and escalation. Name the facilities and IT responders, their handoff points and the authority to shut down compute before thermal limits are reached.
- Record recovery criteria. Specify the flow, temperature, pressure and alarm conditions that must be stable before compute load is restored.
What operators should do during a power outage
The exact sequence belongs to the site’s emergency operating procedure, but a practical decision path is:
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →- Confirm the electrical state. Check utility loss, UPS status, ATS position, generator status and the power feeds serving the cooling chain.
- Confirm coolant flow. Read CDU pump state, supply and return temperatures, pressure and flow. Do not assume a powered display means the pump is moving coolant.
- Confirm heat rejection. Check facility-water temperatures and pump status, or the air-conditioning equipment used by a liquid-to-air CDU.
- Check automatic failover. Verify that the standby pump or CDU is running and that the control system reports the expected active path. Investigate a failover alarm rather than repeatedly resetting equipment.
- Protect the thermal margin. If flow or heat rejection is unstable, follow the approved sequence for workload migration, load shedding or server shutdown. Avoid starting additional compute while the cooling path is uncertain.
- Maintain communications. Keep facilities, IT operations and incident leadership on the same status record, including temperatures, alarms, power source and actions taken.
- Escalate equipment-specific faults. Use the CDU and facility-plant manuals for valve positions, isolation and restart instructions. Do not apply a procedure from another vendor’s system.
Restoring cooling after power returns
Utility restoration is not the same as thermal recovery. A site may remain on generator power, or equipment may have transferred to a standby path that must be inspected before returning to normal.
Best Value
- Data Center Coolant
- 25% Inhibited Propylene Glycol
- JeffCool ISF 25
- High thermal conductivity
- Verify stable utility or generator power and the intended ATS position.
- Confirm that primary and redundant pumps, controllers and facility-side equipment report healthy status.
- Check for trapped air, abnormal pressure, leaks or unexpected valve positions according to the manufacturer’s procedure.
- Allow supply and return temperatures to stabilize while monitoring flow and alarms.
- Restore compute in stages, watching the cooling response at each load increase.
- Close the incident only after telemetry, alarms and redundancy have been confirmed, not merely after servers boot.
How to compare cooling resilience designs
When evaluating a new installation or an expansion, compare the entire failure path rather than counting backup devices.
| Evaluation question | Evidence to request |
|---|---|
| Backup scope | One-line electrical diagrams showing UPS, generator and alternate-feed coverage for CDU and facility equipment |
| Runtime and capacity | Declared UPS and generator loading, autonomy assumptions and the loads excluded from backup; a universal runtime is not stated by the available guidance |
| Redundancy | Pump, CDU, control-panel, valve and heat-rejection redundancy, including shared points of failure |
| Failover | Automatic sequence, transfer alarms, interlocks and results from witnessed failover tests |
| Monitoring | Flow, pressure, temperature, leak and power telemetry with an independent alarm path |
| IT protection | Documented thresholds and automation for workload reduction and graceful shutdown |
| Recovery | Vendor-approved restart order and measurable conditions for returning compute load |
ASHRAE emphasizes fault-tolerant CDU design and backed-up power; Vertiv emphasizes coordinated system design, automated failover and leak detection that can support controlled shutdowns. Those principles are useful evaluation criteria, but no single source establishes a ranking of all liquid-cooling systems.
Common misconceptions
“Liquid cooling means the servers are safe when the power fails.”
Liquid is only the heat-transfer medium. Pumps, controls and heat rejection still require power, and the servers themselves may remain energized.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
“A UPS on the rack protects the CDU.”
The rack UPS may support IT equipment while the CDU, facility pumps or chiller are on a different electrical bus. Protection must be traced across the whole chain.
“Redundant pumps eliminate outage risk.”
Two pumps do not protect against a shared controller, valve, pipe, power feed or facility-plant failure. Redundancy helps only when the alternate path is independent and tested.
“The system will always shut down automatically.”
Some designs have load-shedding or shutdown integration; others rely on operator action. Confirm the actual control sequence and alarm ownership in the site procedure.
Quick Recap
Questions to ask the cooling and facilities teams
- Which exact devices must remain powered for coolant circulation and heat rejection?
- What happens if utility power, a CDU pump, a controller or the facility chiller fails individually?
- Which transitions are automatic, and what alarm confirms that the standby path is carrying the load?
- What measured condition starts IT load reduction or graceful shutdown?
- How are leaks, low flow and rising temperatures detected when the primary monitoring network is unavailable?
- What checks must pass before compute load is restored after an outage?
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.
