AI Data Centers Strain Power Grids and Water Supplies Across U.S.
The physical infrastructure demands of artificial intelligence are creating resource bottlenecks, forcing communities to confront who pays for the boom.
The artificial intelligence revolution depends on a sprawling physical infrastructure that is colliding with the hard limits of America's power grids and water systems. As tech giants race to build data centers, communities are discovering that the cloud has a very terrestrial cost.
Northern Virginia hosts over 300 data centers that handle an estimated 70% of global internet traffic, according to a report by Ken Silverstein in Forbes. The region's transmission grids are so congested that new facilities face wait times of up to seven years for grid connections. The strain has sparked organized resistance: in central Virginia, farmers and conservationists are mobilizing against the Valley Link Project, a proposed $1 billion, 115-mile transmission line with 165-foot towers designed solely to power the data center industry.
Why it matters
The resource demands of AI infrastructure are no longer abstract. They are forcing real tradeoffs between economic competitiveness and local resource constraints. Every megawatt consumed by a data center is power unavailable for manufacturing, housing, or healthcare facilities. As global data center energy consumption approaches 1,050 TWh by year-end—equivalent to Germany and France combined—the question is not whether to build AI infrastructure, but who bears the cost and on what terms.
The water problem hiding in plain sight
Data center advocates frequently compare their water use to golf courses or farms, but this comparison obscures a critical distinction. Agricultural operations typically use untreated water from rivers or wells. Data centers consume treated, potable municipal water—the same supply residents drink from their taps.
In Northern Virginia, potable water use by data centers more than doubled between 2019 and 2023, reaching nearly 2 billion gallons annually, Silverstein reported. Texas faces a projected demand of 399 billion gallons by 2030. UC Riverside professor Shaolei Ren notes that AI cooling relies on "blue water"—clean, fresh water from rivers and lakes—which evaporates and is permanently removed from local communities, unlike "green water" naturally held in soil for agriculture.
The renewable energy illusion
Tech companies claim to power their operations with renewable energy, but this assertion ignores grid physics. Wind and solar are intermittent sources that only generate power when weather permits. Data centers require constant, 24/7 baseload electricity. When renewables go offline, utilities must maintain fossil fuel plants as backup.
Luke Metzger, Executive Director of Environment Texas, told FOX7 that an estimated 130 new gas-powered plants have been proposed for Texas, partly to meet data center demand. "If they're all built, that's going to have as much climate pollution as 27 million cars," Metzger said.
Because battery storage is not yet scalable, regional utilities keep aging coal plants running or construct new natural gas peaker plants to prevent blackouts. A company's renewable energy certificate does not change the physical reality: their round-the-clock power demand locks fossil fuels into local grids.
Communities push back
Virginia recently enacted the nation's first data center electricity tax at 1.1 cents per kilowatt-hour. The levy addresses a stark reality: the power these servers demand threatens to raise residential utility bills by hundreds of dollars annually to fund new grid infrastructure. Rather than blocking development, the tax forces tech companies to pay the true cost of their strain on communities.
The shift from unchecked expansion to resource accountability is accelerating. Tech companies are discovering that while software deploys instantly, pouring concrete and obtaining permits for power lines are stubborn physical realities.
These details were first reported by Ken Silverstein in Forbes.
This is an original analysis by the Omega editorial team. Source reporting: AI Watch.
Want systems like this working for your business?
Book a Call

