Finance, AI & Market Briefings

FERC's summer assessment quantifies the adequacy picture regionally rather than as a single national headline - which is where the analytically useful content lives. The Mid-Atlantic BPS, which includes most of PJM, is rated adequate under normal conditions but carries elevated risk under extreme conditions, particularly given the concentration of new large loads in Northern Virginia. New England faces a structurally tighter margin, with ISO-NE's reserve margin already below its 15% target and natural gas supply constraints that worsen under cold snaps. Those are different risk profiles driven by different mechanisms, and conflating them into a single 'summer adequacy' narrative misses the operative variable.

The probabilistic methodology FERC staff uses is worth understanding, because it shapes what 'adequate' means in the report's terms. The assessment uses Loss of Load Probability and Expected Unserved Energy as reliability metrics, calculated by running a Monte Carlo simulation over a large ensemble of weather scenarios, demand realisations, and resource outage probabilities. A region passes the adequacy threshold if it meets the one-day-in-ten-years LOLP criterion across the simulation ensemble. The critical assumption embedded in that criterion is that weather, demand, and resource outages are drawn from independent distributions. When they are correlated - as they are during prolonged heat events that simultaneously increase air-conditioning load, reduce thermal plant efficiency, and stress natural gas supply infrastructure - the tail probabilities can be substantially larger than the Monte Carlo criterion implies.

Concentrated large loads create a specific technical challenge for this methodology. Traditional demand-side variability is geographically diffuse and weather-correlated in predictable ways: hot days in Northern Virginia drive cooling load up proportionally across a large number of small commercial and residential customers. A 500 MW data centre cluster in the same geography adds a load component that is weather-independent - its compute load is driven by inference demand and training schedules, not temperature - but concentrated enough that a single interconnection failure or local transmission constraint creates a step-change in the local load balance. FERC's Monte Carlo model, calibrated on historical data without this load profile, will systematically underestimate the probability of coincident events in regions with high data-centre concentrations.

The natural gas linkage adds a correlation pathway that standard LOLP models handle imperfectly. Gas-fired generation provides roughly 40% of PJM's summer capacity. In extreme heat events, gas demand for cooling and industrial processes competes with gas demand for power generation on the same pipeline infrastructure. Pipeline constraints can cause gas-fired generators to derate or trip at exactly the moment peak power demand is highest - introducing a supply-demand correlation that classical probabilistic adequacy models, which treat fuel availability as exogenous, do not capture. FERC notes this risk qualitatively; quantifying it requires coupled gas-power flow modelling that most RTOs do not run operationally.

The practical implication for infrastructure investors is that the standard adequacy metric - reserve margin against peak demand - is a less reliable guide to actual risk in a correlated-stress environment than it was in the historical period it was calibrated on. A region with 15% theoretical reserve margin but high gas dependency, high data-centre concentration, and ageing transmission infrastructure has a materially different tail-risk profile than a region with the same headline margin but a more diverse resource mix and distributed load. FERC's summer reports are valuable precisely because they show the regional distribution of these risk factors, but they are not designed to quantify tail risk in the correlated-stress regime - which is where serious reliability modelling now needs to go.

The policy consequence is that summer reliability reports are becoming de facto pre-approval documents for large-load interconnection requests. State commissions in Virginia and Maryland already require data-centre developers to demonstrate that their load addition does not materially worsen regional reliability metrics. FERC's assessment provides the reference baseline against which those demonstrations are evaluated. Once that feedback loop is established - adequacy report shapes procedural obligations for the load categories that drove its stress findings - the time from final investment decision to energisation for a large data-centre project will include a reliability impact study as a critical-path item.