Abraham Silverman is a research scholar at Johns Hopkins University and principal of Silvergreen Energy Consulting. Suzanne Glatz is principal consultant at Glatz Energy Consulting and was previously a director with PJM’s System Planning Division. Mahala Lahvis is a student at Johns Hopkins University.
Grid operators in the United States are on a building spree. Billions are being invested in electrical infrastructure, in part, driven by an unprecedented expansion of hyperscale data centers and the return of manufacturing to the United States.
Regulators are asking increasingly urgent questions: should federal and state policymakers seek to insulate mom and pop retail customers from the impacts of data center expansion, and if so, how can regulators ensure fair and equitable treatment of all customers? These issues were front and center at the Federal Energy Regulatory Commission’s Nov. 1 meeting on data centers and co-location of data centers at existing generating facilities.
The concept of creating a protective moat to ring-fence retail customers from risk has been around for generations. Federal and state regulators should unite around a modern-day data center ring fencing movement to protect small retail customers from footing the bill for transmission upgrades needed to serve big data centers and ensure that data centers contribute their fair share to the fixed costs of operating the transmission grid.
The risk of under-investment in the transmission grid & stranded costs
In the United States, the dominant method for allocating the cost of new transmission facilities necessary to serve new load customers is to spread the costs across all retail customers — a practice known as socializing or peanut-buttering costs. New factories and other large users of electricity make substantial commitments to the local community and operate for long periods of time, paying enough in transmission charges to pay off the utility’s investment in new transmission facilities. Adding new customers to the system reduces transmission rates for all users of the system — since the total cost of operating the transmission grid (the numerator) is spread over more sales (the denominator), resulting in a lower $/kilowatt-hour charge.
Data centers, however, challenge this conventional wisdom in several ways. First, there are concerns over whether data centers will contribute sufficient revenues to pay for the transmission facilities that are being built on their behalf. The investment necessary to serve these new mega-loads is large and the payback period is long — that is, a new data center customer must pay for transmission service over many years to justify the large upfront investment. Data center customers, however, can evaporate as quickly as they’ve appeared. If the anticipated data center load does not materialize, or if they close shop prematurely, the transmission expansion costs can be stranded. Instead of payments from the data center going to pay off the transmission investment and funding a portion of the costs of running the grid, remaining customers would be left paying for a more expensive grid without offsetting revenues.
Second, data centers, in their quest to come online as quickly as possible, are increasingly seeking to co-locate with already existing generators. Colocation has the potential to shatter the existing grid funding paradigm because, as a number of commenters noted at the FERC technical conference and in comments filed on Dec. 9, generators pay the actual costs of connecting to the grid but are exempted from purchasing additional transmission service and thus make significantly smaller contributions to the fixed costs of operating the grid.
Notably, these dynamics can play out in a variety of ways in addition to colocation — including if the data centers end up not using the amount of electricity that they predict or end up leaving the grid after only a few years or significantly reducing their consumption, whether because of evolving business needs, overbuild of data center capacity, an unexpected downturn in data needs, or simply increased energy efficiency as technology evolves and each server rack consumes less energy.
Neither state nor federal regulators can do it alone
While the Federal Power Act of 1935, in theory, established a bright line between federal and state regulation of the electric power grid, the line often gets smudged in practice. When they are at their most effective, federal and state regulators often work together to protect retail customers, a philosophy amorphously referred to as cooperative federalism. Given the ability of data centers to shift costs across multiple states, a cooperative approach is desperately needed. Neither regulator can do it alone; state regulators have jurisdiction over retail sales of electricity, while federal regulators oversee transmission spending.
The need for federal-state coordination is perfectly illustrated in an ongoing proceeding before the Ohio Public Utilities Commission. There, Ohio utilities are battling over whether data center customers should be required to enter into long-term contracts before being allowed to connect to the grid. While long-term contracts, backed by postings of financial security may protect the customers of a single utility, such payments do little to protect customers in the rest of PJM.
FERC and state regulators have famously worked together to solve tough issues in the past, including the Joint Federal-State Taskforce on Electric Transmission, and FERC 2006’s small generator interconnection standards rulemaking, where FERC harmonized “state and federal practices by adopting many of the best practices … recommended by the National Association of Regulatory Utility Commissioners.”
We suggest several immediate avenues for FERC-state cooperation.
First, federal-state consensus is needed around whether to ring-fence non-data-center retail customers from bearing the costs of expanding the transmission grid to meet data center load. Doing so would require changes to the generations-old paradigm that new load connection costs should be socialized based on the theory that new load leads to lower consumer costs.
States cannot do it alone. Transmission necessary to serve data centers cannot be contained to a single service territory or even a single state and quickly spill over into neighboring states. The concerns over data center cost allocation in fact echo concerns heard from red state regulators about being forced to bear the cost of blue state environmental policies which have become a flashpoint in FERC’s recent Order No. 1920 long-term transmission planning rule. FERC Commissioner Mark Christie issued an extensive dissent focused on the impact of one state’s policies on consumer costs in other states. That same dynamic is playing out here where pro-data center policies in one state are driving costs for neighbors.
Second, FERC and state regulators should cooperatively address the proverbial elephant in the data center — whether it is unduly discriminatory to apply new rate structures exclusively to data centers. Treating data centers differently than other large loads may split regulators, creating a significant risk of confusion as different states and utilities adopt different approaches.
FERC recently punted on this issue in several cases — including the Susquehanna nuclear-Amazon Web Services colocation case and another involving Basin Electric Power Cooperative, where FERC rejected a proposal to adopt differentiated rules for crypto mining load (data centers’ alter ego) as unduly discriminatory. A federal-state consensus whether data centers are similarly situated to other large loads would go a long way to resolving this tension.
Third, state and federal regulators must address colocation. The retail sale of electricity in a colocation configuration is subject to state regulation. However, transmission rates and cost allocation, established at the federal level, are equally implicated. Grid-connected data centers pay for transmission service based on their gross load, while colocated data centers pay for transmission service based on their net load. If data centers increasingly co-locate with generation, it could potentially starve the grid of critical revenues and resources, leaving other retail customers to pick up the financial slack.
One potential solution to the colocation issue is for federal and state regulators to agree that data centers need to pay transmission and distribution costs based on their gross load, without regard to whether they are located with generation or not. This type of coordinated approach would ensure that wholesale and retail rates are set in a comparable manner and minimize incentives to play regulators off each other.
Fourth, we need to adopt standardized load interconnection processes that could be used by both state and federal regulators to ensure that duplicative data centers aren’t adding to the already staggering rate of load growth. States could use their authority over new load requests to ensure appropriate reporting of load and avoid double-counting of new data center needs.
Successful integration of these new giants into the transmission grid requires a transparent load interconnection queue and rules governing the priority of new loads seeking interconnection service. This would significantly benefit both state and federal regulators, as well as provide additional transparency for developers on what can be a frustratingly vague process. Again, federal and state regulators would benefit from close coordination to do together what neither can do separately.
