The Grid Defection Threat

Electric utilities spent years anticipating that surging data center demand would be a windfall — a massive, reliable new load that would justify transmission investment and help spread fixed infrastructure costs across more customers. That calculus is now being complicated by a structural shift in how the largest data center operators are thinking about their power supply.

Industry disclosures and project announcements suggest that by the end of this decade, a meaningful share of new data center capacity could be built to operate partially or fully independent of the utility grid. Hyperscalers are signing direct power purchase agreements with generators, building on-site generation assets, and in some cases developing dedicated transmission infrastructure that bypasses the regulated utility system entirely.

Why Data Centers Are Pursuing Grid Independence

The motivations are straightforward. The largest data center operators have power demands measured in gigawatts, timelines measured in months, and reliability requirements that exceed what most utility interconnection queues can practically deliver. The average wait time for large load interconnection has stretched to years in many regions, creating a bottleneck that grid-defection strategies are designed to circumvent.

Cost is a secondary but significant factor. Utilities recover their infrastructure investments through rates that are averaged across all customers. A large data center that can source power more cheaply through a direct PPA or on-site generation has a financial incentive to do so — particularly if it can avoid the cross-subsidies embedded in retail rates.

The Stranded Cost Problem

For utilities, grid defection by large industrial customers creates a well-documented financial problem: stranded costs. Utilities have built or planned transmission and distribution infrastructure to serve data center loads. If those loads are partially or fully self-supplied, the infrastructure investment still exists — but the cost must be recovered from a smaller remaining customer base.

Regulators in states with high data center concentration are beginning to grapple with this. The traditional model — in which large commercial and industrial customers effectively subsidize residential rates — works only if those customers remain on the grid. A significant shift toward self-supply by the largest customers could force upward rate pressure on residential and small commercial customers who have no practical alternative.

Planning Under Uncertainty

The planning challenge is equally significant. Utilities model load growth years or decades in advance to determine what generation, transmission, and distribution investments to make. Data center load has been among the most volatile planning inputs in recent memory — projects are announced, delayed, modified, or cancelled on timescales that make it difficult to plan infrastructure with confidence.

Grid-independence strategies add a new layer of uncertainty. A data center that announces a 500MW load connection and subsequently self-supplies 300MW through on-site gas turbines or a dedicated nuclear PPA represents a planning error that utilities may not discover until after making infrastructure commitments.

Policy Responses

Several state public utility commissions are exploring policy responses, including standby charges for customers who maintain grid connections as backup while primarily self-supplying, and requirements for advance disclosure of large load self-supply plans. The Federal Energy Regulatory Commission has also been watching the trend, particularly as it affects the economics of transmission planning under its Order 1920 framework.

The tension between data center operators' need for speed and flexibility and utilities' need for planning stability is unlikely to resolve quickly. The scale of AI infrastructure investment — measured in hundreds of billions of dollars annually — means the stakes for getting the regulatory framework right are enormous for both industries.

This article is based on reporting by Utility Dive. Read the original article.