Who pays for innovation? The Buyer or the Bystander?
Utilities call it growth. Politicians call it progress. Residents just call it unaffordable.
This is a column about technology. See my full ethics disclosure here.
The problem with data centers for some is simple: the costs show up in your light bill. According to the US Department of Energy, data centers consumed approximately 4.4% of total U.S. electricity in 2023, and projections for 2028 put it at 6.7 to 12% of total U.S. energy production.
At one point, the internet felt free and weightless. Google a question, stream a movie, no visible bill attached. But as the saying goes, “there’s no such thing as a free lunch”, bystanders are quickly feeling the costs of AI. But there’s a hidden tax of AI. However, as the usage of AI continues to increase (in fact, this is the fastest adopted technology ever), the demand for megawatts of energy is rising in tandem. And those megawatts come from power plants we all subsidize.
There’s a real impact on the locals regardless of their use of advanced computing. In Georgia, families are already paying $43 more a month than they did in 2023, with regulators admitting 80% of new demand comes from data centers. In Louisiana, regulators openly acknowledged that after 15 years, ratepayers will likely inherit Meta’s long-term costs. Politicians call this progress. Residents call it unaffordable.
What we’re watching isn’t innovation, it’s extraction. Tech companies negotiate in private, utilities expand, and communities inherit the bill.
What rarely makes headlines is that solutions already exist, and some are showing promise. Terraflow is rolling out industrial flow batteries in Texas that are capable of 10-hour discharges, designed to smooth the spikes from data centers. The Department of Energy is pushing operators to co-locate renewables and microgrids, effectively turning data centers into grid assets rather than liabilities. At the infrastructure level, NREL is testing underground thermal storage systems that stockpile cooling during off-peak hours for use when demand is at its highest.
But with all innovations, there’s always a catch. These solutions aren’t cheap, and someone has to foot the bill. Utilities aren’t in the habit of investing unless regulators force them. Which means the deciding factor is political will, not technical feasibility.
The AI boom is a mirror image of the industrial booms that came before it: railroads, steel, and oil. Each of them reshaped infrastructure and passed the costs downstream. The question is whether we repeat that cycle—or demand better.
This is Part 2 of 4, catch up on Part 1, or check out Part 3.