Exelon's data center pipeline falls 40%, complicating the AI power demand thesis
A sharp drop in Exelon's "high probability" data center queue challenges the power demand narrative that has driven energy equity valuations higher.
Exelon reported Thursday (2026-07-30) that its "high probability" data center load fell nearly 40%, to about 11 gigawatts in the second quarter from 18 gigawatts at the end of 2025 — a decline that sits awkwardly against the wall of capital that has flowed into grid-adjacent equities on the premise that AI infrastructure would deliver uninterrupted power demand growth.7
The category matters. "High probability" loads are projects that utilities have assessed as likely to proceed — not speculative inquiries, not early-stage site surveys. A 7-gigawatt drop in a single quarter means Exelon's own filter rejected or deferred a significant portion of what it had previously considered near-certain demand. The company said it has been entering "transmission security agreements" with potential data center customers, a mechanism that ties interconnection to firmer commitments.7
The broader market has been pricing the opposite trajectory. Capital has rotated aggressively into power generation and grid infrastructure plays on the assumption that AI data center buildout would keep electricity demand expanding well ahead of historical norms. Virginia commercial electricity sales rose nearly 30 million megawatthours between 2019 and 2025, according to EIA data, a pace exceeding any state except Texas. That empirical demand growth gave investors confidence that pipeline projections being cited by utilities were solid.5
But pipeline projections and actual interconnection commitments are different instruments. Exelon's transmission security agreements appear designed to distinguish between the two — and the second-quarter figures suggest a meaningful portion of the 18-gigawatt figure from late 2025 did not survive closer scrutiny.7
PJM Interconnection, which serves 67 million people across 13 states and Washington, D.C., saw real-time demand surge to roughly 163 gigawatts during the week of June 29, 2026, as a heat wave collided with existing data center load and pushed the system to near-record levels, OilPrice.com reported. The physical demand stress is real.4 Yet PJM's capacity market may be telling a different story going forward. Capacity prices hit the auction price cap at the most recent base residual auction, and Jefferies equity analysts said Wednesday (2026-07-15) they expect "long-term structural reforms in the [base residual auction] toward a continuing operating cost model with materially lower prices." If Jefferies is right, the capacity price signal that has underpinned much of the power sector equity re-rating may not hold.6
The EIA's Annual Energy Outlook 2026, published in May 2026, projects server energy consumption at standalone data centers rising through 2050, lending long-run credibility to the demand thesis.2 But the NERC Summer Reliability Assessment, published June 3, 2026, flagged that large computational loads create "operational challenges" and cited recent incidents of unexpected disconnections as a risk operators need to manage ahead of peak demand periods. Unexpected disconnections are a reminder that anticipated load and delivered load can diverge sharply, particularly for facilities still in early operational phases.3
Investors have not been slow to commit to the demand thesis. Fluence Energy shares closed at $24.16 on May 8, 2026, up 98.2% in a single week after the company disclosed master supply agreements with two hyperscalers and a record $5.6 billion backlog, Quick Read Capital data showed.1 Capital rotation at that speed tends to price in certainty that underlying demand projections will deliver. A 40% contraction in Exelon's forward queue, in one quarter, does not map neatly onto that certainty.
The transmission security agreement mechanism Exelon is employing could become a useful leading indicator. If other utilities serving heavy data center load pockets begin disclosing similar filters and similar pipeline shrinkage in their own quarterly results, the power sector equity thesis shifts from whether aggregate demand grows to how much slippage occurs between announced intent and firm grid commitment. The next round of utility earnings will show whether Exelon's queue correction is an outlier or the beginning of a broader pattern.7