PJM Reports $23.1 Billion Cost for AI Data Centers | ForkLog
Artificial intelligence is reshaping the U.S. power landscape—and the bill is coming due. A new analysis from Monitoring Analytics, the independent market monitor for PJM Interconnection, estimates that AI-driven electricity demand has added $23.1 billion in costs across the nation’s largest grid operator.
Surge in Costs and a Shifting Capacity Market
The report finds that rapid load growth from AI data centers is already transforming PJM’s capacity market—the mechanism that pays generators to be available in future years. Across three auction cycles covering 2025/2026 through 2027/2028, higher consumption expectations translated into billions in additional revenue for power producers, costs that ultimately flow through to consumers.
In the first quarter of 2026, average wholesale electricity prices in PJM jumped 75.5% year over year to $136.53 per MWh. The steepest pressure came from capacity payments—essentially reservation fees for generation—which the monitor says surged nearly 400%.
Auctions Push Bills Higher Despite Caps
Even with price caps negotiated in an agreement between Pennsylvania and PJM, end-user bills swelled by $13.8 billion over just the last two capacity auctions, according to the analysis. The monitor characterizes the current dynamic as “unique and unprecedented,” noting that the market was designed for incremental, predictable demand—not for massive, near-term interconnections associated with AI compute.
Why AI Demand Is Different
Traditional load growth tends to be gradual and geographically diffuse. AI projects, by contrast, often concentrate very large loads at single sites and require fast-track timelines. That mismatch strains interconnection queues, transmission planning, and the reserve margins that capacity markets are intended to safeguard.
Proposed Policy Shifts for Data Center Integration
To better align AI growth with grid stability and consumer protection, Monitoring Analytics recommends tightening interconnection and operational rules for data centers:
- Link large-scale data center connections to the availability of new generation resources, preventing load from outpacing supply.
- Adopt protocols enabling full-load disconnection during system stress events to maintain reliability.
- Create a dedicated interconnection queue for projects that bring their own generation, expediting self-supplied capacity without crowding out other needs.
Who Pays for the Buildout?
The monitor warns the challenge is systemic. Bitcoin miners historically sought cheap surplus energy and tolerated curtailment, moderating their grid impact. AI facilities, however, demand high reliability and minimal downtime, which can necessitate new generation and transmission. That raises a core policy question: Should the costs of grid expansion be borne primarily by tech developers whose projects trigger the upgrades, or by ratepayers at large?
Industry Moves Signal More Demand Ahead
Momentum isn’t slowing. In May, TeraWulf acquired a site in Eastern Kentucky to develop high-performance computing infrastructure with potential capacity exceeding 1 GW—an example of the scale now in play across the sector.
Bottom Line
PJM’s market monitor is sounding the alarm: AI’s power appetite is already reverberating through capacity prices and consumer bills. Without targeted policy tools—tying interconnections to new supply, enabling emergency curtailment, and streamlining self-supply—costs could continue to escalate as AI deployments accelerate across the grid.
