Affordable Housing and Environmental Sustainability (1 of 2)

The artificial intelligence (AI) boom is often framed as a simple story of innovation. It is also a scramble for power, grid capacity, and, in some places, water. In the United States, data centers consumed roughly 4.4% of all electricity in 2023, and federal estimates suggest that share could climb to between 6.7% and 12% by 2028 (U.S. Department of Energy, 2024). That is not a minor change at the edges of the grid. It means households and ordinary businesses increasingly compete with giant, nonstop (24x7) facilities that can attract dedicated infrastructure, faster upgrades, and favorable treatment from utilities.

Core contrast

That is the central contrast in the AI consumption debate. Residential, commercial, and data center energy use differ not only in scale but also in political and economic power. Homes have smaller, weather-sensitive loads and almost no leverage. Most ordinary businesses are somewhat better positioned, but they still depend on the same local utility constraints. Data centers sit at the other extreme: round-the-clock demand, long-term procurement strategies, and enough scale to influence where new infrastructure is built and who bears the cost.

Energy costs

The clearest imbalance appears on the monthly bill. On average, households pay more per kilowatt-hour (kWh) than commercial customers. In May 2026, national figures were about 18.05 cents per kWh for residential users versus 14.12 cents per kWh for commercial users. Federal forecasts put average residential prices for 2026 at about 18.02 cents per kWh (U.S. Energy Information Administration, 2026a, 2026b). Put simply, the customers with the least negotiating power are already paying the steepest unit rates.

The current AI and data center buildout threatens to widen that imbalance. Utilities and regulators often present grid upgrades as neutral investments in growth. They are not neutral when costs are socialized while benefits are concentrated. In the Pennsylvania-New Jersey-Maryland (PJM) region, the nation’s largest grid operator, the independent market monitor estimated that data centers accounted for 63% of the increase in the 2025/2026 capacity auction, translating to about $9.3 billion in additional customer costs (Monitoring Analytics, 2025). In places such as western Maryland and Ohio, that meant noticeable projected increases in monthly bills for ordinary residential customers. This is the heart of the political problem: families are being asked to underwrite an infrastructure sprint designed largely for hyperscale demand.

For commercial customers outside the hyperscale tier, the outlook is hardly reassuring. Some large users can negotiate better tariffs than households, but many ordinary businesses cannot. They remain exposed to pass-through costs for new substations, transmission lines, and generation built to serve clustered data center growth. The divide is not simply residential versus commercial. It is increasingly everyone else versus the small number of massive customers with the scale to shape the system.

Availability

If cost is the most visible pressure point, availability is the deeper structural one. The core issue is not whether the United States can generate more electricity in theory. It is whether enough power can be delivered where and when it is needed, without privileging the largest and best-connected buyers. Utilities can announce data center deals faster than they can build the transmission lines, substations, dispatchable generation, and interconnection capacity needed to support them.

The growth trajectory is hard to dismiss. The U.S. Department of Energy has indicated that electricity demand from U.S. data centers may be two to three times higher by 2028 than today, and PJM’s long-range outlook similarly projects summer peak demand rising from about 160 gigawatts in 2025 to 253 gigawatts by 2046, with data center expansion as the main driver (U.S. Department of Energy, 2024; PJM Interconnection, 2025). That does not mean homes will go dark overnight. It does mean every open megawatt, every interconnection queue position, and every infrastructure investment becomes more fiercely contested. In fast-growing places such as northern Virginia, that struggle is already visible.

Water is another stress point that rarely receives equal attention. Public reporting citing a Berkeley Lab estimate indicates that U.S. data centers directly used about 17 billion gallons of water in 2023, and that total could rise sharply by 2028, depending on the pace of facility expansion and the cooling systems used (McCauley & Scanlan, 2025). In regions already concerned about water scarcity, the AI buildout is no longer only a power issue. It becomes a local fight over how limited resources are allocated among industry, utilities, agriculture, and residents.

Accessibility

Accessibility sounds abstract, but it is really about who receives usable service on fair terms. For households, access usually means a regulated connection and little else. They cannot buy premium reliability, negotiate custom infrastructure, or demand a faster path to new capacity. They get the system they are given at the price approved for everyone in their class.

Commercial access is uneven, and that unevenness matters. Small and midsize businesses still operate under the constraints of the local utility, while hyperscale data centers can influence utility planning through long-term contracts, tax incentives, campus-scale commitments, and direct engagement with regulators and transmission providers. In practice, the best access to new energy capacity often goes to customers with the largest loads, the strongest balance sheets, and the most predictable demand. That is not a free-market triumph. It is a policy choice about whose growth counts first.

What the comparison suggests

The comparison points to an uncomfortable conclusion. Households usually pay more per unit of electricity and have less ability to secure scarce capacity. Ordinary businesses may pay somewhat less, but many still lack the leverage to protect themselves from congestion, higher capacity prices, and infrastructure delays. Meanwhile, the largest AI and data center customers can negotiate custom arrangements even as their demand pushes systemwide costs higher. If that sounds backward, it is because it is.

The question is not whether data centers use too much energy in a moral sense. The real question is who pays for the wires, substations, generation, and water infrastructure required to serve them, and who gets first claim on scarce capacity while that buildout catches up. If policymakers want the AI economy, they should stop pretending it is costless. The basic rule should be simple: new hyperscale demand should bear far more of the infrastructure and reliability costs it creates, rather than shifting those costs onto households and ordinary businesses. Otherwise, the AI boom will look less like technological progress and more like a subsidy extracted from everyone else’s utility bill.

 References

U.S. Department of Energy. (2024, December 20). DOE releases new report evaluating increase in electricity demand from data centers.

U.S. Energy Information Administration. (2026a). Electricity monthly update; Electric power monthly.

U.S. Energy Information Administration. (2026b, May 12). Short-term energy outlook: Electricity, coal, and renewables.

Monitoring Analytics. (2025, June 3). Analysis of the 2025/2026 RPM base residual auction—Part G—Revised.

PJM Interconnection. (2025, January 24). 2025 PJM long-term load forecast report.

McCauley, P., & Scanlan, M. (2025, August 19). Data centers consume massive amounts of water—companies rarely tell the public exactly how much. The Conversation.

Ardith Collins is the Founder and CEO of HILOA, an IRS-recognized 501(c)(3) nonprofit developing deeply affordable, climate-aligned housing for low- and moderate-income (LMI) families in small and midsized U.S. cities. HILOA is currently leading the Danville Program in Virginia, an approximately 50-unit all-electric community designed to eliminate debt service from underwriting and help LMI households build long-term stability and wealth. Collins brings nearly four decades of experience in nonprofit leadership, affordable housing development, and climate-focused community investment.