How AI and data centers are testing the U.S. power grid
We are seeing two major trends converge in the U.S. The growth in power demand for data centers and the need for a modern grid. Power transmission and distribution grids are enablers of the energy transition. However, the U.S. electric grid is old, strained and adjusting to a changing energy landscape. Now with the advent of artificial intelligence (AI), data centers and their unprecedented power needs are straining limited grid capacity.
In this context, tech companies, utilities and regulators are having to figure out how technology innovation and economic development are delivered, without unduly burdening other consumers through higher electric rates. This has come to a head in the Midwest, where a local utility and Big Tech firms are in a dispute before the State’s Public Service Commission to decide the electric pricing structure (or tariff) for data centers. The outcome of this dispute may serve as an indicator of things to come across the U.S.
Data centers require significant amounts of reliable power for their operations. A large, hyperscale data center runs continuously and adds the equivalent of 400,000 electric vehicles to the power system1. In its local service territory, the Midwest utility provider estimates that power needs will more than double by 2030 and thereafter increase nearly ten-fold if all interested data centers are developed. To provide this amount of power reliably, utilities must make significant investment in the power transmission and distribution infrastructure. The local utility states that billions of dollars in grid infrastructure are needed to meet the projected additional load for data centers in its service territory.
New electric infrastructure is paid for over time through the electricity rates of those customers that drive the infrastructure need. However, given the scale of power demand, the utility provider is proposing a new tariff specifically for data centers, to ensure long term commitments to the projected electricity demand. Without these commitments, the utility provider fears the demand may not materialize and the utility will be stranded with unnecessary and costly infrastructure upgrades which will need to be paid for by other customers.
Data centers currently fall under general tariffs for large industrial users. The newly proposed, industry-specific tariff is highly unusual. However, because the power demand is so great, the utility provider is seeking a data center tariff with longer contract terms, higher minimum demand charges and greater financial assurance provisions than required of other general customer classes.
The technology firms which own these data centers assert the approach is misguided and unfair. The tariff singles out one industry, rather than applying to all those with similar characteristics, such as loads and usage patterns. Moreover, these firms argue that the approach will stifle innovation and economic growth and will harm data-driven critical functions upon which society relies.
The foundational elements of this dispute are shown in the table below. Since the initial filing, the parties have closed the gap between their positions but remain committed to the fundamental principles underpinning their rationale. On the one hand, the utility is seeking to protect other customer classes and reduce unnecessary investment by ensuring longer term commitments. While on the hand, Big Tech is looking to establish tariffs that drive innovation and growth through the appropriate grid investments without individual industries being singled out.
The core of the dispute:
Utility wants… | Big Tech wants… | |
Applicable industries | …an industry specific tariff for new data centers with a monthly demand greater than 25 MW. | …a tariff that follows established regulatory policy and classifies customers according to their load characteristics rather than by industry type: specifically, a tariff for all electricity-intensive customers with new load greater than 50 MW. |
Contract terms | …contracts not less than ten years with penalties (i.e. exit fees) for leaving earlier. | …options for contract lengths – five-, seven-, or nine-year terms –with lower or no exit fees. |
Minimum demand charges | …minimum billing demand charges that will not exceed 85 percent of contracted capacity. | …similar demand charges that don’t exceed 85 percent, but with options for lower demand charges for longer contact terms. |
These negotiations between the utility, Big Tech firms and other rate payers could have historic ramifications:
- The resolution may set a precedent for other regions where there are similar issues involving data center development, power needs and infrastructure.
- This matter also highlights the growing tension between Big Tech and state regulators. Data center developers wield significant socio-economic and political influence, generating employment and using their data to underpin the operations of critical functions, such as hospitals.
As demand for data centers and accompanying power continue to increase exponentially, regulators will need to manage this growth in a manner that is sustainable and equitable for the data centers themselves, the utilities and infrastructure involved, as well as other ratepayers.
Finally, this debate also highlights the strategic importance of the power transmission and distribution grid. The power grid not only serves as the backbone of our everyday lives, but it is also the enabler of the energy transition. It connects not just data centers, but also heat pumps, electrolyzers (for green hydrogen production), electric vehicles, offshore wind and other technologies that will serve our energy needs of today and tomorrow.
What happens in the matter of this case will have strategic implications locally and nationally: signaling how utilities, regulators and large energy users may collaborate to meet the energy needs of the future.
1IEA World Energy Outlook 2024 Launch Presentation