Data Center Power and Cooling: How It Works

Data center power and cooling refers to the systems that supply electricity to servers and remove the heat they generate, keeping equipment within safe operating temperatures. You enter your facility’s IT load (in kilowatts) and the power usage effectiveness (PUE) to estimate total facility power draw and annual energy cost. The calculator below gives you those numbers, along with the cooling load implied by your PUE.

Formula

Total facility power = IT load × PUE

Cooling power = IT load × (PUE − 1)

Annual energy cost = IT load × PUE × 8760 × electricity rate

Where:

• IT load (kW) – the average power drawn by all servers, storage, and network equipment.
• PUE (dimensionless) – Power Usage Effectiveness, the ratio of total facility power to IT power. A PUE of 1.0 means all power goes to IT; 2.0 means half is overhead.
• Electricity rate ($/kWh) – your blended cost per kilowatt-hour, including demand charges if applicable.
• 8760 – hours in a non-leap year.

Worked Example

Using the default values (IT load = 100 kW, PUE = 1.5, rate = $0.12/kWh):

1. Total facility power = 100 × 1.5 = 150 kW
2. Cooling power = 100 × (1.5 − 1) = 50 kW
3. Annual energy cost = 100 × 1.5 × 8760 × 0.12 = $157,680

The calculator shows 150 kW, 50 kW, and $157,680, matching the steps above.

Why Power and Cooling Are Inseparable

Every watt of electricity that enters a server eventually becomes heat. A server drawing 500 W produces roughly 500 W of thermal output. That heat must be removed or the internal temperature rises, leading to throttling, crashes, or permanent component damage. Cooling therefore consumes a significant fraction of total facility power—often 30–40% in a typical air-cooled data center. The two systems are linked: more efficient cooling reduces PUE, which lowers both energy bills and the environmental footprint.

Power Delivery Path

Electricity flows from the utility through several stages before reaching a server:

• Utility feed – high-voltage (e.g., 13.8 kV) incoming power, often with redundant feeds.
• Transformers – step down to distribution voltage (e.g., 480 V).
• UPS (Uninterruptible Power Supply) – provides battery backup for short outages and conditions power quality.
• PDU (Power Distribution Unit) – steps down to 208/120 V and distributes to racks.
• Rack PDU / busway – delivers to individual servers, often with metering.

Each conversion step loses a small percentage as heat. UPS efficiency typically ranges from 94–98% in modern double-conversion units. Redundancy (e.g., N+1 or 2N) adds more equipment and slightly lowers efficiency, but is required for high availability.

Cooling Methods

Cooling removes heat from the data hall and rejects it to the outside environment. The main approaches are:

• Air cooling (CRAC/CRAH) – Computer Room Air Conditioners or Air Handlers blow cold air through raised floor or overhead ducts. Cold air enters the front of racks, hot air exits the rear. This is the most common method for existing facilities.
• Liquid cooling (direct-to-chip) – Coolant circulates through cold plates attached to CPUs or GPUs. It handles much higher heat densities than air (over 100 kW per rack) and reduces fan energy.
• Immersion cooling – Servers are submerged in a dielectric fluid that absorbs heat directly. It offers the highest cooling efficiency and density but requires specialized hardware.
• Free cooling – Uses outside air or water when temperatures are low, bypassing mechanical refrigeration. This is effective in cool climates and can dramatically reduce PUE.

Most modern facilities use a combination: air cooling for general loads and liquid cooling for high-density GPU clusters.

How to Choose the Right System

Selecting a power and cooling architecture depends on several factors:

• IT load density – Average kW per rack. Above 20–30 kW/rack, air cooling becomes impractical; liquid cooling is often required.
• Availability requirements – Tier III or IV facilities need redundant power paths and cooling equipment, increasing cost and complexity.
• Climate – Cooler regions can leverage free cooling for many hours per year, reducing energy use.
• Budget – Liquid cooling has higher upfront cost but lower operating cost at high densities.
• Existing infrastructure – Retrofitting a raised-floor air-cooled hall for liquid cooling is challenging and may require new piping and rack layout.

For most small to mid-sized deployments, air cooling with a PUE of 1.3–1.6 is cost-effective. For AI workloads with high-power GPUs, direct-to-chip liquid cooling is becoming standard.

Common Problems and Misconceptions

One common misconception is that lower temperatures are always better. In reality, ASHRAE recommends a supply air temperature range of 18–27°C (64–80°F) for most equipment. Running colder wastes energy and provides no reliability benefit. Another issue is hot spots: even if the average room temperature is fine, a single rack with poor airflow can overheat. Proper hot/cold aisle containment prevents recirculation.

Other frequent problems include:

• Humidity control – Too dry causes static discharge; too humid causes condensation. Recommended range is 20–80% relative humidity.
• PUE measurement errors – If you don’t meter the IT load accurately, your PUE calculation is meaningless.
• Overcooling – Setting thermostats unnecessarily low wastes energy without extending equipment life.
• Inadequate redundancy – A single cooling unit failure can cause a shutdown if there is no backup.

Cost Considerations

Energy is the largest operating cost after staff and space. A 1 MW IT load at PUE 1.5 running 24/7 consumes about 13.1 GWh per year. At $0.12/kWh, that’s over $1.5 million annually. Reducing PUE from 1.5 to 1.2 saves roughly $300,000 per year for the same IT load. However, efficiency improvements often require capital investment—such as adding variable-speed fans or liquid cooling—so payback periods should be calculated.

Note: electricity rates vary widely by region and time of day. This calculator uses a flat rate; for more accuracy, incorporate your actual tariff structure.

Safety and Environmental Considerations

Data centers involve high-voltage equipment and large cooling systems. Key safety points:

• Only qualified personnel should work on electrical systems; lockout/tagout procedures are mandatory.
• Cooling systems use refrigerants under pressure; leaks must be handled by certified technicians.
• Battery rooms (for UPS) need ventilation to prevent hydrogen accumulation.
• Water-based cooling systems require leak detection and proper drainage to avoid water damage to electronics.

Environmentally, data centers are significant energy consumers. Many operators purchase renewable energy or use water-efficient cooling towers. The industry is moving toward more sustainable designs, but the core challenge remains: reliably dissipating heat while minimizing energy and water use.

Frequently Asked Questions

What is a good PUE for a data center?

A PUE below 1.2 is considered excellent, 1.2–1.5 is typical for modern facilities, and above 2.0 indicates significant inefficiency. The industry average has been improving, but it varies widely by location and cooling method.

Can I cool a data center with outside air?

Yes, in many climates. Direct air economization brings outside air into the data hall when temperature and humidity are within acceptable ranges. This can eliminate mechanical cooling for most of the year in cool regions, but requires filters and careful humidity control.

How much power does a typical server rack use?

It varies greatly: a standard rack with 10–20 servers might draw 5–15 kW, while a high-density GPU rack can exceed 50 kW. The average is around 7–10 kW, but AI workloads are pushing densities much higher.

What is the difference between CRAC and CRAH?

CRAC (Computer Room Air Conditioner) uses a compressor and refrigerant, similar to a home AC. CRAH (Computer Room Air Handler) uses chilled water from a central plant. CRAH units are generally more efficient and common in larger facilities.

Does liquid cooling really save energy?

Yes, because it removes heat more directly, reducing the need for air movement and lower-temperature refrigeration. At high densities, liquid cooling can cut cooling energy by 30–50% compared to air cooling, and it also allows higher processor performance without thermal throttling.

How do I calculate the cooling load for a data center?

A rough estimate is that cooling load equals the total heat generated by IT equipment plus heat from lighting, people, and external sources. For a quick estimate, use the calculator above: cooling power = IT load × (PUE − 1). For exact sizing, consult a mechanical engineer who will perform a detailed heat load analysis.

Related

• 720p to 1080p Conversion: What It Means
• How to Clear Samsung Mailbox Email, Voicemail & Storage
• Blox Fruits Codes for Devil Fruits?

in it | IT load (kW) | kW | 100 | 0
in pue | PUE | - | 1.5 | 1 | 3
in rate | Electricity rate ($/kWh) | $/kWh | 0.12 | 0 | 1
out total | Total facility power | kW | it * pue | 2
out cool | Cooling power | kW | it * (pue - 1) | 2
out cost | Annual energy cost | $/year | it * pue * 8760 * rate | 0