PJM Report Warns That Data Center Load Growth Has Emerged as Key Variable of Capacity Market Instability and Power Cost Increases
The explosive growth of the AI industry is no longer just a matter of semiconductor and cloud costs. The question is now moving to electricity. MMU, the independent market monitor of PJM — one of the largest electricity markets in the United States — identified data center load growth as the key variable of electricity market instability in its Q1 2026 report. While the report evaluated PJM's energy market as having operated competitively in Q1 2026, it analyzed that the capacity markets for 2025/2026, 2026/2027, and 2027/2028 were not competitive.
The core is clear. Data centers in the AI era are not simply industrial infrastructure. They are emerging as a new demand shock that shakes electricity market prices, transmission network investment, capacity procurement, and general consumer rates simultaneously.
PJM is a large-scale electricity market operator covering the eastern and parts of the midwestern United States. As of March 31, 2026, PJM has an installed capacity of 184,191 MW and is responsible for the electricity market and transmission network operations of a region encompassing some or all of 13 states, Washington D.C., and more than 67 million people.
The message of the "2026 Quarterly State of the Market Report for PJM: January through March" published by Monitoring Analytics can be summarized in three main points. PJM's real-time load in Q1 2026 increased 3.1% compared to the same period of the previous year. The real-time load-weighted average LMP rose 67.8% from $52.20 per MWh to $87.57. Total wholesale power costs jumped 75.5% from $77.78 per MWh to $136.53.
On the surface this is an electricity price increase. But within it is a more structural change. Data center demand is simultaneously shaking two premises of existing electricity market design — demand predictability and the speed of supply expansion.
The report designates data center load growth as the main reason for recent and future capacity market conditions. In particular, it identified data center growth as the core cause of overall forecast load increases, tight supply-demand balance, and high prices. The analysis is that without the actual and anticipated growth of data centers, the same level of supply-demand pressure and high prices in the 2025/2026, 2026/2027, and 2027/2028 capacity market auctions would not have appeared.
The problem does not end with data centers consuming a lot. The part the report points out more strongly is the uncertainty of forecasting. While evaluating that the accuracy of data center load forecasts is highly questionable, it warned that if unconfirmed forecast loads are included in the market clearing process, costs could be passed on to all customers.
The figures are also heavy. According to the report, with existing and forecasted data center loads included, total revenues in the 2025/2026, 2026/2027, and 2027/2028 BRA capacity markets are analyzed to have increased by more than $23.1 billion. Also, in the two most recent auctions, inclusion of data center load was presented as increasing customer billing by $13.768 billion.
This shows that AI infrastructure expansion can ultimately translate into someone's electricity bill. If data centers are capital expenditures for cloud companies, the costs in the electricity market can flow into a structure shared by regional consumers, industrial users, and all power purchasers.
An important distinction in this report is between energy markets and capacity markets. The energy market is a short-term market for actually buying and selling electricity. The report evaluated PJM's Q1 2026 energy market results as competitive. Although prices rose, the causes are interpreted as reflecting short-term market conditions including fuel costs, transmission constraints, emissions costs, and scarcity pricing.
In contrast, the capacity market is a market for securing future electricity supply capability. This market deals not simply with whether electricity exists today, but whether sufficient generation resources are prepared to handle peak demand years from now. The report evaluated the 2025/2026, 2026/2027, and 2027/2028 capacity market results as not competitive. In particular, it pointed out that the capacity shortfall of 208.7 MW in the 2026/2027 BRA expanded to 6,516.6 MW in the 2027/2028 BRA.
This is the key point. AI and data centers are not just increasing today's electricity consumption but are driving up the costs of securing future electricity supply capacity. In other words, data centers are both electricity-consuming customers and a stress test for the design of electricity markets.
In Q1 2026, PJM's total wholesale power costs increased $58.75 per MWh compared to the same period of the previous year. Among this, energy costs rose $42.90 per MWh, and capacity costs rose $14.21. In particular, capacity costs increased 398.1% compared to the same period of the previous year. In the composition of total wholesale power costs, energy 71.5%, capacity 13.0%, and transmission 13.8% accounted for 98.3% of the total.
This is why data center demand is important in electricity markets. Data centers concentrate intensively in specific regions, require high-density power around the clock, and have higher requirements for electricity supply stability than general demand. Such loads simultaneously pressure transmission network bottlenecks, reserve power procurement, and peak response capability — beyond simply being above-average demand.
The core solution presented by the report is clear. Large new data centers must secure their own new generation sources. The report explains this as the 'Bring Your Own New Generation' — BYONG approach. If a data center wants to newly use power, the structure is that it must also bring new generation resources corresponding to that demand.
MMU recommended excluding data center loads from the next BRA and establishing long-term contracts between data centers and generators to stabilize market price formation and distribute long-term risks fairly among participants. This is essentially a signal that the responsibility for AI infrastructure expansion should be placed on the entities creating that demand, not distributed across all electricity market participants.
The direction of the AI era's electricity bill problem is gradually becoming clearer. The decisive factor in where costs go, how risks are distributed, and who bears the burden of the energy transition will be determined by policy decisions, regulatory frameworks, and the formation of market rules. The PJM report demonstrates at least one thing clearly: the AI infrastructure investment cycle has arrived at the power grid level. The era of AI infrastructure not passing through electricity bills has already ended.
