More Than 30 States Want Data Centers to Pay for the Grid They're Breaking

Something unusual is happening in state legislatures across the country: bipartisan agreement. More than 30 states have proposed or implemented measures requiring large-load customers — primarily AI data centers — to pay additional tariffs or shoulder the infrastructure costs their facilities impose on the electrical grid. The policy wave arrives as a new ITIF report attempts to cool the temperature on the data center debate, arguing that the problem is frameworks, not facilities. But for grid operators watching their interconnection queues stretch to absurd lengths, the policy response may already be overdue.

The Shadow Grid Problem

The term "shadow grid" has moved from industry jargon to policy shorthand. At least 46 data center projects with a combined planned capacity of 56 gigawatts are building or planning dedicated power infrastructure — natural gas turbines, battery arrays, and in some cases nuclear microreactors — that operates alongside but largely independent of the public grid. The drivers are pragmatic: interconnection queues in major ISOs now stretch years, and some projects that secured land and financing long ago remain stalled waiting for grid connections, transformers, and generation capacity that does not yet exist.

The result is a parallel energy system emerging in real time. Companies building AI infrastructure argue that going off-grid, at least initially, bypasses those bottlenecks and avoids straining an already stressed system. Critics counter that off-grid facilities externalize costs — they still need road access, water, and emergency services funded by local ratepayers — while contributing nothing to grid reliability.

The Policy Response

State-level responses fall into three broad categories. The first is direct tariffs: surcharges on large-load interconnections designed to recover the cost of grid upgrades that benefit data center operators. The second is cost-sharing mandates requiring data centers to fund transmission and distribution upgrades proportional to their demand. The third, and most contentious, is capacity reservation fees — charges levied on data centers that reserve grid capacity but use it intermittently, effectively blocking other customers from access.

Several prominent tech companies — including Microsoft, Meta, OpenAI, and Amazon — have attempted to get ahead of the regulatory wave by signing the "Ratepayer Protection Pledge," a voluntary agreement to secure their own power and pay for any grid infrastructure their facilities require. Whether voluntary pledges satisfy state regulators remains an open question.

The ITIF Counterargument

A report published April 6 by the Information Technology and Innovation Foundation pushes back on the prevailing narrative. ITIF argues that data centers are not the primary driver of rising electricity demand, that widely cited figures on their grid impact rely on interconnection queue data that overstates real capacity needs due to speculative filings, and that restricting one category of demand rather than improving the interconnection process itself is the wrong policy response.

The report makes a fair point about queue bloat — many filed interconnection requests will never materialize. But that argument is cold comfort to PJM Interconnection, which has warned of potential supply shortfalls of up to 60 gigawatts and strained capacity that could lead to blackouts as soon as 2027. When the largest grid operator in the country is sounding alarms, the debate over whether data centers are the primary cause becomes somewhat academic.

What Industrial Operators Should Watch

The data center tariff debate matters beyond the tech sector. Manufacturing facilities, logistics hubs, and industrial parks that share transmission infrastructure with data centers are already seeing the effects: longer interconnection timelines, rising capacity charges, and competition for transformer equipment with lead times measured in years. If state policies succeed in forcing data centers to internalize their grid costs, it could ease pressure on industrial ratepayers. If the policies drive more facilities off-grid, the shadow infrastructure will keep growing — and the questions about who pays for shared systems will only get harder.

The next twelve months will determine whether the U.S. energy policy framework can adapt fast enough to accommodate AI infrastructure without breaking the grid that everyone else depends on.