Utility ETFs Are Betting on Data Center Load. Ratepayers Are the Collateral.

A Yahoo Finance article this week pitched three utility ETFs as AI-data-center plays, calling the buildout the largest since the 1970s and urging investors to buy now.^[1] The piece is correct about the scale: Grid Strategies reports national load forecasts have roughly quintupled, driven by hyperscaler demand. But the article never asks who pays for the new gas plants, transmission lines, and reserved capacity if those data centers never materialize.

The answer is your bill. Special contracts between utilities and hyperscalers are routinely filed with the economics redacted, while the costs of new generation and transmission are socialized across all customer classes. The Harvard Electricity Law Initiative has documented this repeatedly: existing tariff structures let utilities extract profits from the public to serve big tech, with the burden of proof placed on ratepayers to intervene. When a data center signs a 10-year contract for 100 MW but the utility builds a 40-year gas plant, ratepayers eat the stranded cost for three decades if the load underperforms.

XLU, the largest utility ETF, puts 58% of its assets in just ten megacaps including NextEra, Southern Company, and Duke.^[3] Those are exactly the utilities filing rate cases that cite inflated data-center pipelines to justify capital expenditures they wanted anyway. The EIA projects US capacity rising 50-90% by 2050, but that forecast is only as good as the data-center pipelines feeding it.^[1] And those pipelines are now a rate-case weapon: announced MWs are often speculative, double-counted across territories, and never netted out for historical realization rates. The question every state commission should ask: What share of your forecast is contracted with binding minimum-demand ratchets, and who bears the cost if it doesn't show up?

The clean solution set is known. Virginia's GS-5 tariff and Ohio's AEP deal include minimum terms of 10-14 years, demand ratchets of 85% for transmission and 60% for generation, collateral of roughly $1.5 million per MW, and 100% cost responsibility for dedicated network upgrades. Those provisions isolate the data-center class so residential and small-business rates are not subsidizing hyperscaler capacity. Yet many utilities push weak tariffs that quietly drop the ratchet or the collateral, leaving ratepayers holding the bag.

Meanwhile, independent studies have found that the existing grid can absorb new load if it agrees to be flexible. Duke's Rethinking Load Growth work showed the US system could handle roughly 76 GW of new load with just 0.25% curtailment of annual hours, rising to 126 GW at 1% curtailment. The Camus/encoord/Princeton follow-on turned this into a business model: bring-your-own-capacity plus flexible grid connections can energize a data center 3-5 years sooner than a firm hookup while sparing ratepayers roughly $764 million per GW in supply costs. That option is almost never studied in the dockets where utilities claim new gas is essential for reliability.

Investors buying utility ETFs are betting that the buildout happens and the costs are recovered. Ratepayers are betting that the load shows up and the capacity isn't stranded. The asymmetry is the story. The protective tariff, with high minimum ratchets, cost isolation, and additionality requirements, is the exit ramp. The docket window is open. The question is whether regulators will demand it before the next rate case closes.