DATA CENTER
A facility packed with servers, networking gear, storage, cooling equipment, and power systems.
AI is one workload among many, but large AI clusters can demand extraordinary amounts of power.FIELD GUIDE · AI, POWER & THE GRID
ARE AI DATA CENTERS REALLY THAT BAD?
They want gigawatts, water, land and new power lines. Here’s what the AI boom actually costs a community — and what it may give back.
UPDATED 2026-09-21
NO PHD REQUIRED
ELI5
An AI data center is a building full of computers that run and train models. The computers use electricity; cooling them can use more electricity and water; and the local grid has to deliver enormous amounts of power reliably. The argument is not whether data centers use resources. They do. The fight is over scale, location, who pays for new infrastructure, and whether the benefits justify the costs.
LOAD
How much electricity a customer draws from the grid at a given time.
A giant new load can require new generation, transmission, substations, or all three.PUE
Power Usage Effectiveness compares all facility electricity with electricity used directly by computing equipment.
Better cooling and power systems can shrink overhead, but they do not erase the energy used by the chips.WATER USE
Some cooling systems consume water directly, while electricity generation can also have a water footprint.
The local impact depends heavily on cooling design, climate, and the power mix.WHY THIS BECOMES A FIGHT
WHY ARE PEOPLE FIGHTING ABOUT THIS?
The same server farm can look like industrial growth to one neighbor and a giant new claim on the local grid to another. The benefits can be national or corporate while the transmission lines, water demand, noise, tax deals, and rate decisions are intensely local. That distribution problem is why this turns into a moral argument so quickly.
GET THESE OFF THE TABLE
THE STRAW MEN
The bad arguments first. Nobody gets to win by beating these.
THE BOILING-OCEAN VERSION
“THE SERVERS ARE BOILING THE OCEAN SO BILLIONAIRES CAN MAKE CAT PICTURES.”
THE JUST-BUILD-IT VERSION
“IT'S JUST ANOTHER FACTORY. BUILD THE POWER PLANT AND QUIT COMPLAINING.”
NOW MAKE THE GOOD ARGUMENT
STEEL MAN THE CASE
Give the people you disagree with the version they would actually defend.
THE SKEPTICAL CASE
THE COSTS ARE REAL AND LOCAL
Data-center electricity demand is growing fast enough to change grid planning. New infrastructure can take years, water stress is location-specific, and communities reasonably want guarantees that large industrial customers pay their share instead of shifting costs onto households. Efficiency gains matter, but total demand can still rise faster than efficiency improves.
A more efficient furnace still burns more fuel if you build ten times as many furnaces.
THE BUILD CASE
THE GRID CAN GROW, AND THE LOAD CAN PAY FOR IT
Electricity demand is not automatically environmental damage. New load can support investment in generation and transmission, data centers can be flexible about location and timing, and their economic value can justify infrastructure expansion. The relevant question is whether new supply is clean, reliable, and financed so existing customers are protected.
A new city needs roads and power too. The argument is who builds them and who gets the upside.
FOLLOW THE MONEY
WHO PAYS? WHO WINS?
HOUSEHOLDS
They lose when utilities socialize infrastructure costs that should have been assigned to unusually large new customers. They can benefit if new investment expands supply and the rate design protects ordinary customers.
DATA-CENTER OPERATORS
They gain access to scarce power and locations. Their real contribution depends on what they pay for interconnection, generation, transmission, water, taxes, and long-term grid upgrades.
LOCAL COMMUNITIES
They host the physical footprint and may receive construction work, tax revenue, and infrastructure. They also bear the most visible land, noise, water, and grid consequences.
THE REST OF US
We get the AI services, cloud capacity, and economic output even when the physical costs occur somewhere else.
RECEIPTS, NOT VIBES
WHAT DO WE ACTUALLY KNOW?
U.S. DEMAND IS ALREADY LARGE
Lawrence Berkeley National Laboratory's 2025 update estimates data centers could account for 11.8% of U.S. electricity use in 2030, with a scenario range of 9.5% to 15.3%.
That is a system-scale planning problem, not a rounding error.GLOBAL DEMAND IS EXPECTED TO MORE THAN DOUBLE
The IEA projects electricity generation serving data centers rising from about 460 TWh in 2024 to more than 1,000 TWh in 2030 in its base case.
Renewables provide nearly half of the additional supply in that case, but gas, coal, and nuclear also contribute.THE RANGE IS WIDE BECAUSE THE FUTURE IS NOT FIXED
DOE's 2024 report projected U.S. data-center electricity at roughly 6.7% to 12% of total demand by 2028. Newer estimates moved again as planned capacity and AI hardware changed.
Any single scary number should be treated as a scenario, not destiny.WHAT WOULD SETTLE SOME OF THIS?
WHAT WOULD CHANGE THE ARGUMENT?
- Utility tariffs that clearly show whether very large data-center customers are paying the incremental generation, transmission, and interconnection costs they cause.
- Measured local water consumption and watershed stress for specific facilities, rather than national averages applied everywhere.
- Evidence that new data-center load is accelerating genuinely additional clean generation and transmission instead of prolonging higher-emission supply.
- Long-run local data showing whether promised tax revenue, jobs, and infrastructure benefits materially outweigh subsidies and public costs.
TAKE THIS TO DINNER: The fight is not whether data centers use resources. It is who pays for the next megawatt, gallon, substation and acre — and who keeps the upside.
RECEIPTS
The guide is the map. These are the sources behind the substantive claims.