What People Are Wondering

No. Individual use has negligible impact. Training a GPT-4-class model can use as much electricity as 1,000 U.S. homes for a year. All the ChatGPT queries in a day might equal a few homes. Your personal queries are not the problem — and stopping wouldn't change the infrastructure dynamic at all.

The issue is systemic: how data center infrastructure gets built, who pays for it, and whether it's designed to benefit communities or just extract from them. That's a policy and governance question, not a personal consumption one.

Search "[your utility name] rate case 2025-2026" to find public filings. PJM's Independent Market Monitor attributed 63% of the 2025/2026 capacity auction price increase — $9.3 billion in additional costs — to data center load. Across three consecutive auctions, data center-related costs totaled $21.3 billion, or 45% of all capacity costs.

These are public proceedings. Consumers can submit comments. Most states have consumer advocate offices that publish plain-language analyses and argue on behalf of residential customers. Finding yours is one of the most effective things you can do.

The permanent job numbers are smaller than industry presentations suggest. Virginia Economic Development Partnership data shows roughly 1 permanent job per $54 million invested in data centers for 2020–2025 — 168 times more expensive per job than other industries. A typical 250,000-square-foot facility creates about 50 permanent operations jobs alongside 1,500 construction positions.

Good Jobs First found that data center developers have received more than $1 million in state subsidies per permanent job created. Nearly half of state data center subsidy programs don't even require job creation as a condition.

The economic value is real — it comes primarily through tax revenue, not employment. Communities that understand this distinction negotiate better deals.

Tax revenue can be substantial. Loudoun County, Virginia collected $875 million in data center tax contributions in 2024 — 38% of county tax revenue from just 3% of land. That funds schools and reduces residential property taxes.

But the full calculation is more complicated. Virginia's sales tax exemption for data centers cost $1.6 billion in foregone state revenue in FY2025 alone. When $4+ billion in grid costs are socialized to all ratepayers, the net public benefit is far less clear than industry presentations suggest. Communities deserve full-cost accounting — not shell companies, code names, and NDAs.

A two-tier grid means some regions get modern, reliable infrastructure while others stay on equipment from the 1960s. If private investment only flows to areas where data centers want to build, tech corridors get 21st-century power while rural and lower-income communities are left maintaining aging systems alone.

This is why federal policy matters. The DOE's grid transmission goals and GRIP Program direct investment toward underserved communities, not just data center corridors. The goal is a grid upgrade that lifts everyone — not a private infrastructure network that leaves the public grid further behind.

Both, with real proof points. Google DeepMind's system achieves a consistent 30% reduction in cooling energy across Google's own facilities. Its partnership with PJM cuts interconnection approval timelines from years to months. Stanford spinout GridCARE freed up 80 MW of capacity for Portland General Electric using AI to find hidden grid flexibility.

A Yale analysis suggests that if AI grid optimization recovers 12–15% of currently wasted grid energy, it could theoretically offset AI's own demand growth. The question is whether that intelligence stays proprietary or gets built into the public grid infrastructure.

Communities have the right to reject projects. $64 billion in data center projects have been blocked or delayed since mid-2024, with 25 projects cancelled in 2025 alone. Over 188 organized opposition groups are active across 40 states — and opposition is bipartisan: 55% Republican, 45% Democrat among opposing elected officials.

When a community says no, the data center usually relocates to a jurisdiction with fewer protections. Communities that engage early typically secure better outcomes: negotiated setbacks, noise limits, water restrictions, local hiring commitments, and community benefit agreements. Engagement beats rejection for producing lasting change.