What's Going On in PJM

PJM is asking for extra operating flexibility in the middle of October, clearing future capacity at historically high prices, and studying an enormous pipeline of proposed power plants and batteries. Those developments belong to the same story: the grid needs usable supply sooner than the investment, construction and rulemaking processes are delivering it.

The immediate pressure comes from warm weather overlapping with equipment maintenance and transmission constraints. The longer pressure comes from expected growth in electricity demand, especially data centers. Between them sits a financial question: how to pay for enough additional capacity without assigning the costs of speculative growth to customers who did not cause it.

PJM coordinates the high-voltage grid and wholesale electricity markets serving about 67 million people in parts of 13 states and Washington, D.C. Its actions reach well beyond a trading screen: they shape utility supply costs, power-plant investment and whether a new industrial site can get dependable electricity. This account uses documents available October 15, 2026, with operational data frozen through October 14.

October usually gives the grid room to work. Generators and transmission owners take equipment offline for repairs, inspections and construction between the summer cooling season and winter heating season. PJM places the main fall work window from mid-September through early December. A system scheduled around mild weather has less flexibility when temperatures rise. PJM's October 12 advisory describes that overlap and a recall of generation maintenance outages effective October 14.

The numbers show the tradeoff. In the public data behind Ask the Grid, the October 13 daily maximum of preliminary hourly demand was 108,109.6 MW, during the hour beginning 5 p.m. Eastern. The same-day generation outage outlook totaled 60,602 MW: 46,430 MW planned, 3,404 MW maintenance and 10,768 MW forced. Planned and maintenance categories together made up 82.2% of that outlook. Scheduled work dominated the reported outage outlook.

Demand had risen 17.8% from October 11, even as the reported outage outlook declined. That combination is more useful than either series alone. A day can have fewer outages than the preceding weekend and still be harder to manage because consumption has risen. Outage recalls can help at the margin; a construction site or a disassembled turbine cannot necessarily return on demand.

These daily figures describe pressure on the system. Calculating an operating reserve margin also requires aligned available capability, imports, exports, operating limits and reserve obligations. Subtracting outage MW from a fleet nameplate number and comparing the result with an hourly load peak would mix incompatible measures. The demand series is preliminary hourly served load; the outage series is a dated availability outlook.

Following the same target date through successive reports reveals something the daily total hides. In its October 8 report, PJM projected 66,667 MW of generation outages for October 14. The October 14 report put that day's total at 61,915 MW, a reduction of 4,752 MW. Read alone, that looks like more room.

Inside that improvement, the planned and maintenance categories fell by 10,550 MW, while the forced-outage component increased from 6,198 MW to 11,996 MW. The 5,798 MW increase offset 55% of the reduction in the other two categories. This is an accounting comparison between outlook vintages: it does not establish that 10,550 MW of equipment returned to service or that every change represents a newly tripped unit. It does show that the composition changed materially as the operating date approached.

The last report-to-report change was particularly concentrated. From October 13's outlook for October 14 to October 14's same-day outlook, forced-outage MW rose 2,575 MW across PJM. 2,092 MW—81% of that increase—was in Mid Atlantic - Dominion, one of PJM's two outage-reporting subregions. The regional total and the system total describe overlapping populations. The outage dataset lets readers follow those vintages separately instead of treating the latest total as the whole story.

PJM's October 14 operations update reports pre-emergency load-management actions on October 13 and 14, covering the afternoon peak and the evening period when solar output falls. Demand-response participants are paid to reduce consumption when instructed. PJM also recalled non-firm sales to neighboring systems. A Maximum Generation Alert gives equipment owners notice to reassess maintenance and testing; a Load Management Alert prepares participants for a possible dispatch. These are distinct steps with distinct consequences. PJM's operating update describes what occurred and when.

The operator records let readers inspect the actions individually: October 14’s afternoon load-management action began at 2 p.m. Eastern across the listed PJM areas; a later action for BGE, Dominion and Pepco began at 4 p.m. Both records show an end time of 7:30 p.m. and preserve the affected areas and source timestamps.

The federal emergency order adds another tool. DOE issued Order 202-26-50 on October 12, effective October 13–16. It directs specified units to operate as needed and permits PJM to call on backup generation as a last resort before declaring, or during, an Energy Emergency Alert 3. In its October 14 operating update, PJM said it had not needed to use large-load backup generation to that point. Establishing the output of specified generators operating under the order requires separate operating evidence. DOE's order and application record define that authority.

Demand response and emergency generation can provide short bursts of flexibility. The practical question is how often PJM needs them, how much they deliver and whether the equipment work they allow to continue is completed before winter. Repeatedly moving necessary maintenance to another day reduces today's exposure while leaving a job on the calendar.

October 13 and 14 look almost identical at the system peak: 108.11 GW and 107.70 GW, both in the hour beginning 5 p.m. Eastern. The location of that demand changed substantially. At that same hour, the Dominion and Mid-Atlantic load areas together rose from 50.95 GW to 55.64 GW. The remaining eight areas fell by 5.10 GW, concealing the 4.69 GW increase in those two eastern areas within a system total that declined just 0.41 GW.

Their combined share of preliminary served load moved from 47.1% to 51.7%. That is a different question for a transmission-constrained system even when the headline demand number barely moves. The comparison identifies where served load changed; it does not attribute that change to data centers, estimate consumption without demand response, or identify which line limited delivery. The hourly load dataset preserves the ten-area breakdown behind the comparison.

PJM's review of two September hot spells identifies transmission limitations around Washington, Maryland and Virginia, including generation bottled behind constraints. Operators also had to cover the evening net-load period, when solar output declined faster than consumption. The system could have substantial generation in aggregate while the particular combination of location, line availability and timing remained difficult. PJM's September operating review connects those conditions to targeted demand response and emergency procedures.

The PJM transmission view puts public transmission lines, substations and zones in geographic context. For the September episode behind this section, our demand-response reconstruction aligns demand, outage outlooks and local prices.

Electricity has to reach the customer through a network with limits. When a line is unavailable, the remaining network carries power differently. An inexpensive generator may be unable to increase output without overloading another facility; a more expensive unit closer to load may have to run instead. That is one reason local wholesale prices can separate even when an entire region clears at one capacity price.

The timing changes the value of new resources, too. Storage can move energy into an evening constraint if it is charged, has enough discharge duration and can deliver at the relevant location. Additional transmission can make remote generation usable. Generation near demand can ease dependence on an import path. Each option addresses a particular constraint; a headline total of proposed MW cannot establish which one is solved.

October 13's verified hourly prices show the financial expression of that geography. For the hour beginning 6 p.m. Eastern, BGE cleared at $2,370.65/MWh, versus $1,770.05/MWh at Western Hub. Of the $600.60/MWh gap, $548.90/MWh—91.4%—came from the difference in congestion components. The energy component was identical; marginal losses account for the remainder, within source rounding. This is a zone-to-hub price comparison, not an attribution to a particular transmission facility.

The day-ahead market had already priced an expensive evening: BGE's price for that hour was $1,316.82/MWh. Real time added $1,053.83/MWh. About 74.1% of that increase came from the common energy component, which rose from $856.83/MWh to $1,638.13/MWh. The remaining change was mainly congestion. Both statements can be true: congestion explained most of BGE's premium over Western Hub, while energy explained most of BGE's increase from day-ahead to real time. They answer different comparisons.

The sequence also explains why watching only the demand peak misses part of the evening. Preliminary hourly load peaked at 5 p.m.; the verified BGE price peaked at 6 p.m.; the day's gas and oil generation snapshots peaked at 7 p.m. Between the 5 p.m. and 7 p.m. readings, solar fell from 6,540 MW to 27 MW, while gas rose from 42,663 MW to 46,355 MW and oil from 3,105 MW to 4,043 MW. Hourly served load declined 1,540 MW over those two hours. The fuel readings are instant snapshots and load is an hourly average, so they are shown separately. The dispatch pattern does not identify DOE-authorized generation or measure how much demand response delivered. PJM's 2 p.m. action and 4 p.m. action both extended to 9 p.m. across their listed areas.

The energy market schedules and prices electricity for particular hours. The capacity market buys an obligation to be available in a future delivery year. The first is measured in $/MWh; the second in $/MW-day. Capacity therefore enters investment and utility procurement decisions before the delivery period begins. An October operating alert and a July auction result can both signal pressure while describing different years and different products.

The RTO capacity price for 2028/29 was $325/MW-day, compared with $333.44 for 2027/28. Both reached their applicable caps. The cap is calculated using the accredited capacity of PJM’s reference resource. Changes in that accreditation alter the cap expressed per UCAP MW; the price movement alone cannot demonstrate an improvement in the supply balance. The delivery year runs from June 1, 2028 through May 31, 2029. PJM's July auction announcement explains the collar and delivery obligation.

A capacity payment supports the economics of keeping or building a resource that can meet a reliability obligation. Construction also needs a credible stream of future revenue, financing, equipment, permits and a connection. A high annual auction price can make a project more attractive while still arriving too late to produce a plant by the required year.

The July procurement left 6,831.3 MW UCAP below PJM's reliability requirement. UCAP is unforced capacity: the amount credited toward reliability after accreditation, rather than a machine's rated output. The gap is a failure to procure the full planning target, which includes reserves. It raises modeled risk; it is not a prediction that this much electricity will be unavailable in every peak hour. PJM's auction report makes that distinction.

Cleared capacity increased 3,733.2 MW UCAP from the preceding auction; PJM attributes that increase mainly to higher accreditation factors and fewer unoffered MW. Most capacity in the auction represents commitments from an existing resource base. New construction takes time, and uprates increase the capability of equipment already in the system. Accreditation changes can also alter the counted supply without an equivalent change in physical equipment. Those distinctions explain why a substantial increase in cleared capacity need not mean an equally substantial wave of completed power plants.

Demand growth is only one side of the balance. In July, PJM's board said about 15 GW of generating resources had retired since 2022. That is a gross retirement figure, not a measure of net fleet change: additions and uprates occur alongside closures. It nevertheless identifies capability that the system has had to replace while preparing for larger loads. The PJM generation directory provides a way to examine individual plants and their technologies. The board's July statement presents retirements and new demand as concurrent pressures.

A replacement also has to perform during the hours that create reliability risk. A solar project's rated output, a battery's discharge power and a gas generator's dependable winter contribution measure different things. Solar depends on daylight and weather; a battery needs stored energy and sufficient duration; a thermal unit needs fuel and functioning equipment. Accreditation attempts to translate those characteristics into a contribution to reliability.

The development question therefore has several parts: how much new equipment is completed, what it can deliver during difficult conditions, whether that output reaches the relevant load and which existing resources remain available. An orderly replacement can combine technologies and transmission. Counting nameplate MW alone skips the performance and timing that make the combination work.

PJM plans against expected future demand. Once a forecast enters the procurement calculation, it can influence how much capacity the market seeks and the price existing resources receive. A project can therefore affect investment signals before consuming electricity. Forecast quality is a present cost question as well as a future engineering question.

The January 2026 forecast projects strong growth in both seasons. It also reduced the 2028 summer peak forecast by 4,414 MW from the prior year's vintage. PJM attributed the near-term reductions to changes in large-load adjustments, electric vehicles and economic assumptions. It required firmer commitments for near-term large-load additions and discounted less certain projects farther out. The same report can show a large long-term expansion and a meaningful downward revision. PJM's January load report documents both.

Monitoring Analytics, PJM's independent market monitor, argues that forecast data-center growth has been central to tight capacity conditions and that the forecasts need stronger substantiation. Its first-half 2026 report estimates $29.37 billion in additional capacity-market revenues across four delivery-year auctions attributable to existing and forecast data-center load. That is the monitor's estimate under its auction-analysis assumptions, using actual clearing prices, quantities and uplift MW. It is not a measured sum of household bill increases or a valuation of the October emergency order. The monitor's report sets out its argument.

The useful test is project-level credibility: an executed service agreement, credible financing, construction progress, equipment delivery and a realistic ramp in consumption. Counting every inquiry as certain demand risks overbuilding and allocating costs to projects that disappear. Excluding credible demand risks building too little. A transparent forecast should let states and market participants distinguish those cases and see revisions as projects advance or withdraw. The PJM large-load directory provides facility profiles to investigate alongside those forecast assumptions.

PJM's reformed interconnection process requires financial commitments and site control to screen projects before studies proceed. Its August announcement accepted 715 projects into the first cycle. By the October 1 snapshot, 617 projects representing 168.6 GW remained in that cycle. Study admission establishes a stage of development; the studies determine network impacts, required upgrades and costs. PJM's process announcement explains the change to first-ready, first-served screening.

Explore the PJM interconnection queue view for project-level requested capacity, technology, status and proposed completion dates. Its current publication can differ from the October 1 snapshot used here; projects are located at county level.

The October snapshot lists 195 GW under study and 49.3 GW with agreements across the reported pipeline. These are different project populations. The larger number cannot be added to accredited auction commitments as if it were operating supply. The smaller number represents an important development milestone, followed by financing, construction, network upgrades and commissioning. PJM's October 1 statistics project final agreements for Cycle 1 in May 2028—close to the start of the 2028/29 capacity delivery year, with construction still to follow where required.

An agreement is valuable because it resolves connection requirements and lets developers make more concrete investment decisions. It does not guarantee that a project will be built. PJM's October 6 Pennsylvania legislative testimony identifies permitting, siting, supply-chain delays, interest rates and infrastructure costs as further obstacles. That account is PJM's explanation of development barriers, rather than an independent allocation of responsibility for each delayed project. The testimony makes clear why queue reform and completed construction must be tracked separately.

There are faster routes, with limits. On October 8 PJM announced its first three expedited-track projects: two batteries totaling 1,660 MW nameplate and a 455 MW gas uprate, planned for mid-2029. The process requires state support for expedited permitting and a minimum reliability contribution. Those projects could improve later supply; their proposed timing does not resolve this October's operating problem or establish a full year of availability beginning June 2028. PJM's announcement preserves that distinction between acceptance and operation.

PJM proposed a separate Reliability Backstop Procurement to attract new capacity. On September 29, FERC accepted it for filing but suspended it for five months, setting a February 28, 2027 effective date subject to refund and further proceedings. FERC raised issues involving cost allocation, transmission-owner exit rules and load-serving-entity collateral, and encouraged a revised filing that could move the process forward. The FERC order is a conditional regulatory step, rather than a completed purchase of supply.

PJM then confirmed that the procurement would not start September 30 and that its timetable would be reviewed. Bilateral matchmaking between new loads and new supply is a parallel route; confirmed contracts can offset procurement needs. The watchpoint is an executable procurement design and resulting commitments, followed by delivery milestones. PJM's September 30 update sets out the delay and the parallel contracting process.

There has since been a concrete procedural step. In an October 12 motion, PJM asked FERC to pause the paper-hearing proceedings while it prepares a revised filing, planned by October 29. It scheduled member and transmission-owner consultation for October 22. Those are a request and an announced filing timetable; they do not establish approval of revised procurement rules.

A second proposal concerned new large loads. FERC's October 9 order accepted a large-load definition and registry but rejected the proposed Interim Resource Adequacy Service, bring-your-own-new-capacity provisions and linked auction demand exclusions. PJM argued that its service would support wholesale reliability while preserving state responsibilities. FERC found that parts of the submitted framework intruded on retail regulation and that other provisions lacked sufficient justification. Consequently, July and August proposals cannot be described as the rules now governing new data centers. The original order in Docket ER26-3515 identifies the provisions that survived.

Long-term revenue certainty can help finance new supply. Someone also has to accept the risk that the demand it was built for arrives late, grows more slowly or never materializes. An allocation tied to a forecast is different from one that follows actual customers over time. FERC's backstop proceeding examines precisely that relationship between a procurement's beneficiaries and its costs.

At the household level, wholesale costs pass through utility procurement and state regulation. New Jersey supplies a concrete historical example: its regulator's February 2025 supply-auction announcement projected $26.87 more per month for a PSE&G residential account using 650 kWh, beginning June 2025, and identified PJM capacity costs as the main driver of the increases. That is one utility, usage assumption and procurement period. It shows the route to a bill; it does not estimate a 2028/29 bill or isolate the effect of a particular data center. The NJBPU announcement publishes the assumptions and utility comparisons.

Governance determines who can propose changes to that machinery and how quickly they proceed. On September 30, the Members Committee endorsed the members’ amended term sheet, according to PJM’s voting record. The package addresses filing rights, state participation and stakeholder procedures. PJM’s agenda posted October 13 schedules an October 16 review of governing-document revisions to implement the endorsed provisions. Member endorsement is a step toward rule changes; the resulting documents and applicable approvals determine what takes effect.

The next useful evidence has to match the timescale of the problem. For this fall: equipment returning from maintenance, the evening availability of supply, regional transmission constraints and delivered demand response. Lower demand alone would show a weather reprieve; more dependable capability at the same demand level would show additional operating room.

For the next capacity delivery years: updated forecasts with transparent project assumptions, the next auction's accredited supply and reliability requirement, an executable backstop process and evidence that contracted resources can meet their dates. A capped auction price needs the quantity result beside it.

For the buildout: projects moving from study to agreement, from agreement to construction and from construction to commercial service, together with the transmission needed to deliver their output. Storage duration and accreditation, generator performance, retirement dates and large-load commissioning all affect the balance. The test is usable capability arriving when and where demand materializes.

To follow developments after this reporting cutoff, open PJM Radar for current grid findings and the PJM Energy Wire for market reporting and filings.

The October chart uses public PJM data accessed through Ask the Grid: hourly preliminary load and generation outage outlooks. The window is October 1–14 in America/New_York. All 14 days have 24 valid hourly load observations. The system total is derived from ten load areas; each daily peak is the maximum hourly average. The outage series uses the report date equal to the target date for PJM RTO, with stored revisions collapsed. These frozen extracts were retrieved October 15 and are subject to subsequent source revision.

The 82.2% figure is (46,430 + 3,404) / 60,602. The 17.8% demand increase compares October 13 with October 11. The capacity gap share is 6,831.3 / 156,012.9; the new-generation-and-uprates share is (317.1 + 207.6) / 138,317.8. Auction prices are nominal RTO clearing prices by delivery year. Forecast bars compare endpoints from one January 2026 vintage; both endpoints are forecasts. Queue counts preserve the October 1 snapshot of a document updated in place. No queued nameplate MW are converted to UCAP.

The additional Ask the Grid analysis holds October 14 constant while comparing seven outage-report dates, October 8–14, for three reported regions. Load-area comparisons use the same UTC hour on October 13 and 14. The price reconstruction joins verified hourly real-time LMPs with day-ahead LMPs by exact pricing-node ID and delivery hour; all six checked locations have 24 observations on October 13. The ten-hour explorer displays noon–9 p.m. The geographic basis uses a same-hour difference; the day-ahead comparison uses the same location and delivery hour across markets. Generation by fuel supplies operating snapshots, not hourly energy. These additional extracts were retrieved October 15 and may change with source revisions.

For the event-level reporting behind this overview, read PJM's September demand-response episode, the July capacity auction and the generators covered by earlier emergency orders. Explore PJM on the map to examine locations and public grid data alongside the regional picture.