The United States is poised for a sharp rise in electricity demand from data centers, yet the number of facilities that will actually be built remains unclear. Proposals for new labs are surfacing while other projects are being cancelled, leaving regulators to grapple with how much power‑generation capacity to fund.
Why data centers are a new kind of load
Traditional large electricity users such as textile mills and refineries required construction timelines that aligned with new power plants, often taking two and a half to three years to become operational. Modern data centers consume a comparable amount of electricity but can be erected in nine to twelve months, forcing utilities to start building gas‑fired generators or solar‑plus‑storage farms a year or more before the data center breaks ground.
During that lead time, advances in artificial‑intelligence hardware and software can alter both the computational workload and the efficiency of the equipment, creating substantial uncertainty about the ultimate power draw of a finished facility.
Cost‑allocation challenges
When utilities over‑build capacity, they risk idle assets; when they under‑build, data centers may compete for scarce electricity, driving up prices for all ratepayers. State regulators must decide which parties—utilities, data‑center operators, or ordinary residential and commercial customers—should shoulder the costs of new generation and transmission.
Regulators typically review utility proposals for reasonableness and public benefit before allowing the expenses to be passed on to customers, but the fast‑moving nature of data‑center projects tests the adequacy of those reviews.
State‑level approaches
Ohio’s major utility AEP employs a “demand ratchet” in its data‑center tariff, billing customers on the higher of current demand or 85 % of the peak demand over the previous eleven months. This mechanism helps ensure that data centers contribute a stable share of the costs even when monthly usage fluctuates.
AEP also requires a credit guarantee equal to 50 % of the expected minimum bill, protecting other ratepayers if a subsidiary‑owned data center were to default on its obligations.
Potential paths forward
Some states, such as Florida, have approved agreements that require data‑center operators to pay for 70 % of agreed‑upon demand regardless of actual usage, further solidifying cost recovery. Others, like Missouri, are experimenting with revenue‑sharing models that return a share of profits earned from flexible large loads to ordinary customers.
These varied experiments illustrate a broader search for a pricing framework that balances the need for reliable electric supply with the financial protection of residential and small‑business ratepayers.
Frequently Asked Questions
What drives the uncertainty in data‑center electricity demand?
Data centers can be built quickly, while power‑plant construction takes longer, and rapid advances in AI hardware and software can change the amount of electricity the facility ultimately consumes.
How are regulators trying to ensure utilities aren’t left with unused capacity?
States like Kentucky are conditionally approving new generators only if utilities can demonstrate a real need, while Ohio uses demand‑ratchet tariffs and credit guarantees to tie payment more closely to actual or expected usage.
What mechanisms exist to protect ordinary ratepayers from the cost of large data‑center projects?
Approaches include demand‑ratchet billing, credit guarantees, agreements that lock in a percentage of demand payments, and revenue‑sharing models that return a portion of utility profits from large customers to the broader customer base.
How should policymakers balance the need for reliable power with the risk of over‑investing in new generation for data centers?
