A data center requesting hundreds of megawatts is not just another customer signing up for service. It is an infrastructure negotiation that strains the entire electricity planning system. The core problem isn't just how much power these facilities consume, but how to make them a permanent enough fixture to justify building new power systems around them.
As AI drives a surge in data center development, public debate has focused on quantities: electricity usage, new generation needs, and rising bills. But the real issue is a shortage of institutional infrastructure. The old forecasting world of gradual demand trends has been replaced by project-level shocks from large, discrete loads that arrive faster than utility planning cycles can accommodate.
The electricity system has a long history with energy-intensive industrial loads. But data centers scramble that model: they arrive in large concentrated increments, care intensely about speed to power, are mobile across utility territories, and are uncertain-some backed by major companies, others contingent on financing or permits.
A new paper on large-load tariffs by Angela Navarro and Molly Knoll for the National Association of Regulatory Utility Commissioners captures the shift. Historical load forecast errors were driven by gradual factors. Today's errors come from localized, infrastructure-intensive projects that are often seeking service faster than utilities can plan.
The underlying difficulty is the phantom load problem. If developers request service in multiple locations before deciding where to build, utilities may see more proposed demand than will ever materialize. PJM has responded by requiring firmer commitments. ERCOT in Texas observed average project delays of 180 days and data centers consuming less than half the capacity originally requested.
The central question is who should bear the risk. When a utility constructs infrastructure to serve a large new customer, it makes investments that are durable, capital-intensive, and difficult to repurpose. This transforms a forecasting problem into a contracting problem. Large-load tariffs can convert forecasts into commitments through long-term contracts, minimum billing obligations, collateral, and exit fees.
AEP Ohio’s experience illustrates this well. After adopting revised data center tariff terms including 25-megawatt eligibility requirements, an 85 percent minimum billing obligation, 12-year contracts, and 36-month termination fees, the interconnection queue fell from 30 gigawatts to 13 gigawatts. The missing 17 gigawatts was not necessarily phantom, but once capacity requests became costly to maintain, the queue began to contain better information.
The larger lesson is that the data center boom did not create weaknesses in electricity regulation. It removed the luxury of ignoring them. Better forecasting tells us which futures are more likely. Better contracting tells us what happens when the actual future is different.