A 100 MW data center sounds like a complete answer. It is not.
The number may describe IT capacity, not the power used by the whole facility. The servers may also run below full capacity. Cooling and electrical systems add another layer of demand.
A useful 100 MW data center power model separates these layers. It moves from maximum IT capacity to average facility load, annual energy use, and electricity cost.
Remember the chain: IT capacity → average IT load → facility load → annual energy → electricity cost.
This article explains the model used in our AI Data Center Power Cost Calculator in Python . No programming knowledge is needed here.
The Whole 100 MW Data Center Power Model
We will use one example throughout the article.
| Input | Value | Meaning |
|---|---|---|
| IT capacity | 100 MW | Maximum power assigned to computing equipment |
| Average load factor | 0.85 | Average IT load as a share of maximum IT capacity |
| PUE | 1.30 | Facility energy divided by IT equipment energy |
| Operating time | 8,760 hours | Hours in a standard 365-day year |
Step 1: Understand What 100 MW Means
MW means megawatt. It measures power at a moment in time.
In this model, 100 MW is the maximum electrical capacity assigned to servers, storage, and networking equipment. It is IT capacity.
It is not yet the average load. It is not annual electricity use. It is not the total facility demand.
100 MW answers: How much IT power can this model support at one moment?
Step 2: Apply the Average Load Factor
Computing equipment does not have to draw its maximum power every hour. We therefore use an average load factor.
A load factor of 0.85 means that the average IT load is 85% of the 100 MW capacity.
100 MW × 0.85 = 85 MW average IT load
This is an electrical assumption. It is not the same as saying that every GPU runs at exactly 85% utilization. Workload, hardware, and electrical draw are related, but they are not identical.
Step 3: Use PUE to Include the Whole Facility
Servers are not the only equipment using electricity. A data center also needs cooling, fans, pumps, UPS systems, power conversion, lighting, and control equipment.
Power Usage Effectiveness, or PUE, connects IT energy with total facility energy.
PUE = total data center energy ÷ IT equipment energy
A PUE of 1.30 means that the facility uses 1.30 units of energy for every 1.00 unit used by IT equipment.
We apply that ratio to the average IT load:
85 MW × 1.30 = 110.5 MW average facility load
The difference is 25.5 MW. That power supports cooling and the rest of the facility.
PUE is normally measured over a period of time. In this simple model, we use one average PUE to estimate average facility demand.
Step 4: Know the Difference Between MW and MWh
MW measures power. MWh measures energy used over time.
The difference is similar to speed and distance.
| Electricity | Travel analogy | Question answered |
|---|---|---|
| MW | Speed | How fast is electricity being used now? |
| MWh | Distance | How much electricity was used over time? |
If a facility draws 110.5 MW for one hour, it uses 110.5 MWh. If it maintains that average for a year, we multiply by 8,760 hours.
Step 5: Calculate Annual Data Center Electricity Use
110.5 MW × 8,760 hours = 967,980 MWh per year
We can also write this as:
967,980 MWh = 967.98 GWh ≈ 0.97 TWh.
The units grow with the number:
- 1 GWh = 1,000 MWh
- 1 TWh = 1,000 GWh
The 100 MW data center power model has now moved from capacity to annual energy.
Step 6: Turn Energy into Electricity Cost
Suppose electricity costs $70 per MWh.
967,980 MWh × $70/MWh = $67,758,600 per year
Under these assumptions, annual electricity cost is about $67.8 million.
This is not a quotation for a real project. Electricity contracts may include demand charges, taxes, time-of-use prices, transmission charges, and other costs.
The model is useful because every assumption is visible.
Read the Full Equation from Left to Right
100 MW × 0.85 × 1.30 × 8,760 hours = 967,980 MWh
IT capacity × load factor × PUE × time = annual electricity use
Each term performs one job:
- 100 MW sets maximum IT capacity.
- 0.85 converts capacity into average IT load.
- 1.30 adds cooling and facility overhead.
- 8,760 hours converts average power into annual energy.
Change any input and the result changes. That is why the model is useful.
Three Common Mistakes
1. Treating 100 MW as total facility demand
In this model, 100 MW is IT capacity. The facility uses more after PUE is included.
2. Treating 100 MW as annual energy
MW is power. Annual energy must be expressed in MWh, GWh, or TWh.
3. Treating load factor as GPU utilization
The load factor is a simplified electrical assumption. It does not directly describe every chip or workload.
How the Python Calculator Uses the Same Logic
The Python calculator does not introduce a different model. It follows the same chain.
average_it_load
= it_capacity × load_factor
average_facility_load
= average_it_load × pue
annual_energy
= average_facility_load × hours
annual_cost
= annual_energy × electricity_pricePython makes the calculation repeatable. It also lets the reader change PUE, electricity price, or load factor and see the result immediately.
Continue with the full calculator: AI Data Center Power Cost Calculator in Python: PUE, Load Factor, and Grid Delay .
The Number to Remember
A 100 MW IT data center does not simply use 100 MW. Under an 85% load factor and a PUE of 1.30, it averages 110.5 MW at the facility level and uses about 0.97 TWh per year.
Once this chain is clear, the Python code becomes easier to read. More importantly, it becomes easier to question.
Is the load factor reasonable? Is PUE likely to change? Is the electricity price fixed?
A good model does not hide these questions. It gives each one a place.
Key Vocabulary
IT capacity
The maximum electrical capacity assigned to computing equipment in the model.
Load factor
Average IT load divided by maximum IT capacity.
PUE
Total data center energy divided by IT equipment energy.
MW
A unit of power at a moment in time.
MWh
A unit of energy used over time.
Related Articles
- Why Power, Not Chips, May Limit the AI Data Center Boom
- AI Data Center Power Cost Calculator in Python: PUE, Load Factor, and Grid Delay
References
- U.S. Department of Energy, Best Practices Guide for Energy-Efficient Data Center Design .
- International Energy Agency, Energy and AI .
Published: July 2026 · Model assumptions are illustrative and are intended for education and early screening.