How Much AI Is Actually Draining Resources
Some of it's hype, but a lot of it IS reality.
AI continues to creep into more corners of everyday life, and so do the warnings: that it could wipe out our jobs, make us less creative, or unleash something far more unsettling:
Setting aside the more out-there predictions, there are real, near-term concerns we can measure, especially around resource strain driven by building out the data centers that power AI technology.
Electricity:
1: The U.S. Energy Information Administration says residential power prices rose 6% nationwide year-over-year as of August. But the cost spikes in data center-heavy states were much worse:
Virginia: +13%
Illinois: +15.8%
Ohio: +12%
2: Nearly 6 million households are on the verge of collection notices, according to a new analysis by The Century Foundation, a left-leaning think tank.
Across the country, utility delinquencies jumped 9.7% in one year, with the average past-due balance now reaching $789.
3: According to a report published in June by Monitoring Analytics, capacity costs for PJM — the largest power grid operator in the country — jumped from $2.2 billion to $14.7 billion to $16.1 billion in consecutive yearly auctions.
63% of that spike was attributed to current and forecast data-center demand.
4: Data centers used 183 terawatt-hours of electricity last year, more than 4% of total U.S. demand. By 2030, data center consumption is expected to grow by 133% to 426 TWh.
In Virginia, data centers made up a hefty 26% of electricity demand in 2023.
5: One big caveat on all this is research suggesting AI-driven improvements in energy efficiency “could make up for the additional energy required by data centres.”
Water:
Electricity demand is rising sharply as AI expands. Water use, however, appears far less affected.
Per the A16z newsletter:
Training Grok-4 required ~750 million liters of water.
Sounds big — until you realize a single square mile of U.S. farmland uses 1.2 billion liters annually.
The U.S. has 1.37 million square miles of farmland.
Politics and Policy:
1: Voters are starting to associate AI with higher electric bills — A new Artificial Intelligence Policy Institute (AIPI) survey found:
72% of voters are concerned about AI’s rising energy consumption.
61% support taxing the electricity used by AI companies to help pay for grid upgrades.
Support includes majorities of independents and Republicans.
Only 20% oppose the idea.
2: Electricity and data centers have become ballot box concerns in 2025 races.
Newly elected Virginia Gov. Abigail Spanberger won while calling on data centers in the state “to pay their own way and their fair share” of rising electricity costs.
New Jersey Gov.-elect Mikie Sherrill ran on freezing utility hikes after prices in the Garden State rose nearly 20% year-over-year.
Georgia voters gave the boot to two Republicans from the utility commission after bills rose six times in two years.
In a letter this month, Senate Democrats accused the White House of cutting “sweetheart” deals with Big Tech that have contributed to rising energy costs and forced consumers to subsidize data centers.
3: Even Silicon Valley concedes that meeting AI’s demand will require a far bigger energy footprint, and accompanying environmental consequences are already being felt.
Last year, Google reported its greenhouse gas emissions have increased nearly 50% over five years, with the main culprit being electricity consumption from data centers and supply chain emissions.
Earlier this year, Memphis residents began complaining about the pollution generated by Elon Musk’s xAI data center.
Bubba’s Two Cents
This might be among the first times AI policy has meaningfully intersected with kitchen-table politics but it surely won’t be the last.