How data centres challenge the electricity regulatory model

Demand from data centres and advanced manufacturing facilities has put unprecedented pressure on the US power sector. This is happening at a time when retail electricity prices are rising to help cover utility investments in new generation and grid infrastructure, the cost of managing an ageing grid and new climate risks.

There is an emerging political consensus in the US that data centres and other large-load customers should pay for the infrastructure built to serve them, rather than for other customers to bear the costs. This will be difficult to achieve under current regulatory frameworks, however, and it poses acute challenges for distribution network operators and the deregulated electricity market model, which is popular across the US and OECD countries.

Distribution network operators typically invest in their grids based on the collective demand of all customers. They then design tariffs and set rates for individual customer groups, such as residential and large industrial customers, based on the average cost to serve a customer in that group. The new imperative is to recover costs directly from individual large-load customers, based on the marginal cost to serve them. This represents a radical change to the regulatory compact.

Distribution network operators in the US have been developing new tariffs to do this. They require large-load customers to sign long-term contracts with strict billing terms, making them directly responsible for the cost of the infrastructure built to serve them. But as Dominion Energy, the distribution network operator serving the most data centres in the US, has acknowledged to its regulator, these tariffs cannot eliminate the risk that the large-load customer will walk away, leaving the distribution network operators with stranded assets. In such a case, either other customers or distribution network operators' shareholders would be responsible for covering the remaining cost.

These challenges are compounded for vertically-integrated utilities, who are responsible for electricity generation and transmission, as well as distribution. They face an entirely new credit-risk profile, with a tiny fraction of customers potentially responsible for a majority of their asset costs.

Electric transmission is one of the most costly, but essential, types of infrastructure necessary to serve large-load customers. It is also the most difficult to confine costs to these customers.

Electric transmission is often developed on a regional basis, with costs allocated broadly across geographies and customer classes. This is based on the long-held finding that all customers in the vicinity of transmission benefit from it, including from greater grid reliability and lower prices. Large loads may violate this assumption, however, requiring transmission investments that are tailored to their particular projects, rather than broader economic development.

It would require novel regulatory treatment, including planning and cost allocation methodologies, to confine transmission costs to the large-load customers who cause them and disproportionally benefit from them.

Electricity deregulation has been adopted to some degree across nearly all advanced economies. In this model, electricity generation is separated from transmission and distribution, with competitive generators facing near-term and long-term market prices for power. While considered progressive, this model poses challenges to accommodating large loads.

In a regulated market, vertically-integrated utilities are able to differentiate pricing between customers. They can hold residential and small commercial rates constant, for example, while charging large-load customers based on the costs they incur. This is not possible in deregulated markets, where all customers are exposed to common prices for electricity and (in some cases) generation capacity.

There is a fundamental tension in deregulated markets between grid reliability and the political imperative to shield other customers from the costs of large loads (see Figure 1. below). Distribution network operators consider only transmission and distribution capacity when evaluating a data centre or other large load’s grid connection request. It is up to the invisible hand of the market to ensure that sufficient generation is built to supply the load once it interconnects, otherwise risking blackouts. Such generation is only built if electricity and capacity prices are high enough to incentivize it, raising prices for all customers.

In the most deregulated markets in the US, Texas and the PJM Interconnection, forward prices are far below what would be required to incentivize the construction of new generation facilities. It is, therefore, impossible to accommodate the growth in AI data centres and maintain grid stability without prices increasing for all customers.

As electricity is famously hard to store, electricity markets are more localized than most commodity markets. This means that local politicians are exposed to the fallout from grid unreliability and price increases and can be counted on to take action to prevent it. We have seen early examples in the US states of Pennsylvania and Texas, which have signalled a loss of confidence in deregulated markets due to the factors above. As the infrastructural demands and price impacts of data centres in other states and OECD countries grow, expect to see similar intervention. Politicians may be forced to choose between winning the AI arms race and protecting customer bills.