Digital services
Email, websites, business applications, stored files, and AI tools.
Data centers, explained
Behind every click,
there is infrastructure.
A data center is a facility that houses the computing, storage, and network equipment behind digital services. Here is what that means for a community.
01 / How it works
Useful computing.
Physical systems.
The “cloud” is a service delivered through real facilities. AI is one workload alongside business software, storage, communications, and other digital services.
Computing equipment
Servers process tasks. Storage keeps data. Networks move it.
Supporting systems
Electricity, cooling, power protection, and backup systems support the equipment.
A simplified functional diagram, not a facility blueprint. U.S. DOE · Efficient data center design
Site or campus
A location that may contain more than one building.
Colocation
A facility where customers rent space and supporting infrastructure.
Hyperscale
Very large computing operations designed to scale. Check a source’s definition before comparing counts.
02 / Power explained
Megawatts measure draw.
Megawatt-hours measure use.
A power rating alone does not tell you annual consumption. You need the average draw and the time it runs.
Try a simple example.
Move the slider to change an example facility’s average electricity demand. The calculation assumes that average is sustained for the full period.
1 MW = 1,000 kW1 MWh = 1,000 kWh
This is arithmetic, not an estimate for a real project. A designed connection limit, a peak, and an average draw are different values.
EIA · Energy and peak demand1 MW200 MW
Use arrow keys to adjust the example.
Over 24 hours600 MWh
Over 365 days219 GWh
Energy = average power × hours
25 MW × 8,760 hours = 219,000 MWh/year
03 / Cooling & water
Follow the heat.
Then follow the water.
Computing generates heat that must be removed. Cooling systems differ, so “a data center uses this much water” needs a facility-specific explanation.
At the equipment
Air or liquid carries heat away from servers. The liquid inside a cooling loop can circulate repeatedly.
At the building
Heat still has to leave the facility. Evaporative equipment uses water differently from dry heat-rejection systems.
Beyond the site
Electricity generation can also use water. On-site totals and a full water footprint are different measures.
Design depends on climate, equipment, and operating conditions. U.S. DOE · Efficient data center design · Berkeley Lab · Water use of data center workloads
04 / Check a claim
Look past the headline.
These distinctions make conversations more useful. Open a topic for the context to check.
Does “planned” mean “it will be built”?
No. A proposal, an approval, a building under construction, and an operating facility are different stages. Loudoun County explicitly cautions that not every project in its pipeline will be realized.
Ask: Which stage is this number counting, and when was it updated?
Loudoun County · Data centers by the numbersDo all data centers use the same amount of water?
No. Hardware efficiency, workloads, cooling, local climate, and electricity supply can produce very different water footprints.
Ask: Is the figure measured or modeled? Does it include water used to generate electricity, and what is the local water source?
Berkeley Lab · Water use of data center workloadsDoes “100% renewable” describe every hour?
A renewable energy certificate represents attributes of renewable generation. Certificates may be purchased separately from electricity. An annual matching claim by itself does not establish renewable supply in every hour.
Ask: What is being matched, over what period, and where? Are the certificates retired for this customer?
U.S. EPA · Renewable energy certificates · U.S. EPA · Electricity and unbundled certificatesDoes a lower PUE mean a smaller total footprint?
Power usage effectiveness (PUE) compares total facility energy with IT equipment energy. A lower ratio means less overhead per unit of IT energy. Total consumption can still grow when the computing workload expands.
Ask: What were total annual electricity use, average demand, and peak demand alongside the efficiency ratio?
U.S. DOE · Efficient data center designCan one economic number describe the whole impact?
Separate direct employees from contractor, supplier, and wider spending effects. Separate construction activity from recurring operations. For public finances, distinguish gross tax receipts from costs, incentives, and obligations.
Ask: Which jobs, which years, whose budget, and what assumptions?
See the dated economic measures →Does a forecast tell us exactly what happens next?
No. Berkeley Lab’s national outlook changes across assumptions about equipment shipments, utilization, and chip life. Forecasts are planning tools, not measurements of future consumption.
Ask: What assumptions drive the range, and is it for Virginia, a utility zone, or the whole country?
Berkeley Lab · U.S. energy report, 2025 update05 / For your community
Six questions
for a proposed project.
Use these as conversation starters with planners, developers, utilities, and elected representatives. These are the initiative’s suggested questions.
What exactly is proposed?
How many buildings and phases? What is approved, and what remains uncertain?
Is power ready?
What are the average and peak demands? Which grid upgrades are needed, who pays, and what happens if expected demand does not arrive?
What happens in a drought?
What are annual and peak-day water needs, the supply source, and the operating plan during shortages?
What will neighbors experience?
Ask for sound modeling, nighttime and low-frequency assessment, setbacks, generator testing plans, and a complaint contact.
What benefits are measurable?
How many permanent on-site roles? What are projected tax receipts, public costs, and commitments to local hiring?
What will be reported?
Which commitments will be independently checked? Who publishes results, and what happens if a commitment is missed?
Our policy response
The Virginia Data Center Innovation and Accountability Act is a citizen proposal built around planning, measurement, flexibility, and accountable incentives.
Read the discussion draft →