A data center that doesn't even exist can raise your electricity bill (www.vox.com)

🤖 AI Summary
A rapid surge in AI-driven data center development is already pushing up U.S. electricity costs — not just from facilities that exist, but from ones that are only proposed. Data center counts nearly doubled between 2021 and 2024 and companies may spend $1.6 trillion on U.S. data center hardware by 2030. Grid operator PJM and the DOE now warn data centers could consume a much larger slice of electricity (estimates range from ~9% by 2035 to as high as 12% by 2028). Large hyperscale sites demand hundreds of megawatts, and utilities’ forecasts of future load — often inflated by duplicate bids or tentative proposals — are already provoking record capacity-auction prices and higher monthly bills for households (recent PJM auction prices jumped 22%, adding roughly $16/month for some customers). Technically, the problem is speculative capacity planning: utilities tell PJM how much generation and transmission to add based on anticipated data center loads, so even unbuilt projects drive investment that gets socialized into rates. That can create stranded-cost risk if many projects never materialize. Remedies under consideration include stricter disclosure, more rigorous project review, and financial commitments or deposits from developers to prevent double-counting. If built and managed flexibly, data centers can also stabilize grids and help amortize infrastructure costs — but absent tighter rules, consumers will disproportionately bear the financial risk of an AI-driven power expansion.
Loading comments...
loading comments...