PJM Grid Reliability: the 2027 Shortfall No New Plant Can Fix

The data centres slated to connect to the PJM grid between 2029 and 2030 alone could draw more electricity than the entire state of Pennsylvania consumes today. That is not a projection stretched across a decade. It is two years of new load, added on top of everything else the grid already carries.

That finding comes from an independent study by Synapse Energy Economics, commissioned by the Pennsylvania Public Utility Commission (PUC) and released in September 2026. Its conclusion is blunt: across every scenario the analysts modelled, new in-state generation fails to keep pace with demand, and PJM misses its reliability planning standard every year from 2027 through 2030.

The stress is no longer a forecast. The 2027/2028 capacity auction already cleared at the price ceiling and still came up short of what the region needs, the first such shortfall since PJM’s capacity market launched in 2007.

Here is what the data actually tells you: which generation technologies stand to benefit, where the regulatory constraints are binding, and where investor capital is most likely to find durable tailwinds across the 2027-2030 window. The reliability deficit is not a risk to avoid. For the right asset classes, it is the thesis.

The reliability math that PJM cannot escape

Start with demand, because that is where the pressure originates. PJM’s 2025 long-term forecast projects summer peak load rising by 32 GW, roughly 21%, between 2024 and 2030. Data centres account for approximately 30 GW of that increase, which is 94% of near-term demand growth.

AI data centre energy demand has grown nonlinearly as inference workloads scale alongside training runs, meaning the 30 GW load figure attributed to data centres in PJM’s forecast may itself be conservative if model deployment accelerates faster than planners assumed.

Now hold that against supply. The 2027/2028 Base Residual Auction cleared at the FERC-approved price cap of $333.44/MW-day and still fell 6,517 MW short of PJM’s Reliability Requirement. That was the first capacity shortfall in the nineteen-year history of the Reliability Pricing Model.

Then it happened again. The 2028/2029 auction cleared at the system-wide cap of $325/MW-day, procuring 138,318 MW of unforced capacity (UCAP) plus 10,864 MW from Fixed Resource Requirement entities.

Two consecutive auctions hitting the FERC ceiling is not cyclical tightness. It is the market’s own pricing mechanism signalling structural undersupply, and that signal is worth pricing into any generation asset thesis inside the PJM footprint.

PJM’s reserve target shortfall report on the 2027/2028 Base Residual Auction attributes the capacity gap directly to an unprecedented surge in data centre load, confirming that the pricing ceiling and the physical shortfall share a single structural cause rather than reflecting normal auction volatility.

The trajectory keeps climbing. PJM’s updated 20-year forecast anticipates summer peak demand rising to over 241 GW, averaging 3.6% annual growth across the next decade.

The scale in one sentence New data centres added to PJM between 2029 and 2030 alone could consume more electricity than the entire state of Pennsylvania uses today, according to the Synapse Energy Economics study.

The gap is widening, not self-correcting. That is the foundational frame for every investment decision that follows.

Why the generation buildout is not closing the gap fast enough

The pipeline looks healthy on paper. Gas combined-cycle capacity is projected to rise from 61.4 GW in 2025 to 71.8 GW by 2030. Solar climbs from 14.7 GW to 56.1 GW, onshore wind from 11 GW to 21 GW, and battery storage from 0.5 GW to 7.9 GW over the same window.

Then the constraints arrive, and each one bites on a different timeline.

The gas problem is the one that closes the door on the near-term window. Turbine procurement now runs five to seven years, meaning a unit ordered in 2025 may not arrive until 2030-2032. The 2027 shortfall is effectively already locked in for gas, because the equipment cannot physically be built in time.

Natural gas demand from data centres is compounding the turbine procurement problem, as operators pursuing dedicated gas generation behind the meter are competing for the same equipment and installation contractors as grid-scale developers, adding pressure to a supply chain already constrained by five-to-seven-year lead times.

Meanwhile, PJM reduces the accreditation value it assigns to incremental gas resources, citing winter reliability and fuel supply concentration risk. Adding more gas produces diminishing capacity credit, which weakens the economics of piling into a single fuel.

Solar faces a different ceiling. PJM studies indicate the system can run reliably with up to 30% wind and solar penetration, but only with massive transmission expansion and complementary firm resources. Push solar’s capacity share above roughly 20% without that support and operational violations appear.

At the same time, dispatchable supply is exiting. PJM’s own risk analysis flags 40 GW of dispatchable generation at risk by 2030, including 15 GW of coal tied to EPA rules, with 14-29 GW of coal retirements projected across the region by 2040.

The constraint that matters most for each technology

• Gas: Five-to-seven-year turbine lead times, plus accreditation dilution as more gas is added.
• Solar: Dependence on firm resources and transmission; operational violations above roughly 20% capacity share without them.
• Coal: EPA-driven retirement trajectory removing dispatchable capacity through 2040.
• Nuclear: One-to-two-year PJM interconnection reviews on top of long development timelines, ruling out near-term relief.
• Battery storage: Queue position is the binding factor, though co-location offers a bypass.

The read for investors is a timeline one. The 2027-2030 window was largely determined by decisions made before 2025, so the real capital question is what positions best for 2030-2035.

What data centres are doing to grid operations that the capacity numbers do not capture

Capacity adequacy is a planning metric. It measures whether enough megawatts exist to meet peak demand. It says nothing about what happens second to second when a single load behaves unlike anything the grid was designed around.

Data centres do exactly that. To protect sensitive equipment, they can automatically disconnect during grid disturbances, dropping more than 1,000 MW of computational load off the bulk system in seconds. To the grid, that abrupt drop mimics the failure of inverter-based resources, throwing frequency and voltage out of balance.

This is not a 2027 forecast. It has already happened.

The alert that confirms it arrived In May 2026, the North American Electric Reliability Corporation (NERC) issued a rare Level 3 “Essential Actions” Alert after repeated events of more than 1,000 MW of computational load dropping off the bulk system, sourced to data centre clusters in Virginia and Texas.

A Level 3 Alert is NERC’s most serious tier of essential action, and it was triggered by operational behaviour, not by a capacity shortage. The frequency and voltage problem is present now, ahead of the megawatt deficit still building toward 2027.

Compounding the difficulty is an institutional gap. Data centres are not NERC-registered entities, so the fastest-growing load category on the continent sits outside the framework designed to manage grid reliability. That gap shows up in three ways:

• Registration accountability: Data centres are not directly subject to NERC Reliability Standards.
• Modelling data: Planners lack the operational data they need to model this load accurately.
• Real-time coordination: Curtailment during stress events depends on load behaviour that operators cannot directly command.

NERC expects North American electricity demand to rise by roughly 160 GW by 2030, with about 90 GW driven by data centres. The scale alone would strain any grid; the operational behaviour makes it a distinct risk class.

For investors, the distinction matters. Capacity adequacy risk (the 2027 shortfall) and operational stability risk (the present frequency and voltage problem) are addressed by different assets and priced differently. Generation and storage positioned to provide stabilising services carry a premium the capacity market price alone does not fully reflect.

Pennsylvania’s regulatory response and what it signals for the broader PJM footprint

Pennsylvania is the most capacity-significant state in PJM, and it is moving first. What it builds now is the template other governors will reach for once the reliability gap becomes impossible to ignore in their own states.

The state has assembled a three-part framework around data centre load, layering cost obligations, clean energy mandates, and permitting incentives.

1. HB 1834 (passed the Pennsylvania House 104-95 in March 2026): grants the PUC authority to regulate commercial data centres, mandates security deposits and emergency load restrictions, and sets clean power targets of 10% by 2027, 14.5% by 2030, and 32% by 2035.
2. HB 2650: requires data centre projects to meet demand through incremental capacity and to source rising percentages of electricity from clean, firm resources located within Pennsylvania, matching the HB 1834 timeline.
3. Executive Order: offers preferential permitting to data centres with peak demand over 25 MW that commit to sourcing electricity from new power supplies.

Read together, these are not three separate bills. They form a coherent demand-management strategy that raises the regulatory cost of connecting large loads while steering that load toward clean firm capacity.

How the PUC is operationalising the framework

Legislation sets the direction. The PUC supplies the enforcement teeth.

In June 2025, the PUC voted 3-2 to issue a tentative order proposing a statewide model tariff for hyperscale data centre connections, governing grid access and cost-sharing so existing ratepayers do not absorb the upgrade bill.

Then in September 2026, responding directly to the Synapse findings, the PUC voted unanimously to rewrite emergency electric load control regulations, explicitly defining the curtailment order for data centres during stress events. The shift from a divided 3-2 vote to a unanimous one tracks how quickly the reliability evidence has hardened the regulatory consensus.

Pennsylvania’s proportional share of PJM capacity is projected to fall from 23% in 2025 to 20-21% by 2030 and 14-17% by 2040. That decline, even as the state remains PJM’s top generator, is what makes clean firm capacity built inside Pennsylvania so valuable.

Governor Josh Shapiro has been lobbying counterpart governors to engage more directly with PJM policy, which is the mechanism by which this approach spreads. The signal for investors is twofold: regulatory cost is rising for data centre operators, but the same framework is actively creating preferential channels for developers who can deliver clean firm capacity on the mandated timeline.

Where investor capital is finding durable position in a constrained grid

The tightest grid conditions create the strongest pricing power for the right assets. Here is where the structural constraints open opportunity rather than close it.

• Battery storage gap: The Brattle Group projects PJM needs 16-23 GW of storage by 2032-2040, against roughly 8.1 GW currently active or proposed. That is a structurally underserved market.
• Co-location advantage: Projects using Surplus Interconnection Service to co-locate generation or storage with data centres can bypass the full queue, the single largest timeline differentiator available.
• Clean firm PPA premium: Data centres securing 100% power purchase agreements from incremental clean firm capacity (enhanced geothermal, nuclear uprates) score highest on reliability and decarbonisation, commanding a premium in negotiations.

The storage gap in one line The Brattle Group projects PJM needs 16-23 GW of battery storage by 2032-2040, against roughly 8.1 GW currently active or proposed, representing over $18 billion in projects across the region.

The interconnection queue is the reason co-location matters so much. Timelines to connect to PJM have grown from under two years in 2008 to over eight years by 2025. A project that bypasses that queue holds a compounding advantage as the deficit deepens.

The US battery storage market entered 2026 facing a convergence of IRA tax credit uncertainty, raw material price volatility, and interconnection queue backlogs that determine which projects survive to reach commercial operation in the 2027-2030 window.

PJM’s Fast Lane reforms have begun to accelerate the pipeline, enabling over 4 GW of near-term battery capacity, and the queue freeze is expected to clear its backlog by the end of 2026. Long-duration storage is arriving too, with at least 1 GW of 10-hour assets slated for late-2027 and early-2028 operations.

The gap between 8.1 GW proposed and the 16-23 GW Brattle says the grid needs is growing faster than the queue is clearing. Investors who have already secured queue position hold optionality that appreciates as the reliability deficit widens.

Structural risks that investors must price alongside the opportunity

• Interconnection delay: The eight-year average timeline can strand capital before a project earns revenue.
• Transmission cost uncertainty: Network upgrade liabilities remain unknown until late in development, making renewables especially risky to underwrite.
• Market rule disputes: PJM has resisted FERC Order 2023 study assumptions for battery charging, which storage developers argue artificially prolongs queue processing.
• Speculative load: Many proposed data centre projects may never materialise, creating wide discrepancies in load forecasts and the demand that underpins asset economics.

The opportunity is real, but it is not a free option. Queue position, co-location structure, and long-duration capability are the features that separate durable returns from stranded capital.

What the 2027-2030 window means for energy investors making decisions now

The core asymmetry is this: the 2027-2030 shortfall is structurally locked in. Turbine lead times and eight-year interconnection timelines prevent any new solution from deploying fast enough to close it. That makes the shortage investable, not merely concerning.

The forward synthesis The 2027-2030 reliability shortfall is already locked in by physics and queue timelines. The investable question is not whether the gap exists, but how to position for the 2030-2035 window when new capacity can finally contribute.

The opportunity runs at two speeds. In the near term (2027-2030), advantage sits with assets already in the queue and holding secured interconnection: battery storage with co-location, and gas with existing grid connections. In the medium term (2030-2035), the edge shifts to clean firm capacity, nuclear uprates, and long-duration storage as Pennsylvania’s mandates create preferential demand.

Investors evaluating nuclear uprates as a clean firm capacity play in the 2030-2035 window should read our full explainer on nuclear grid interconnection bottlenecks, which details how transmission constraints and PJM review timelines are limiting deployment even where regulatory approvals have cleared.

The end-of-2026 backlog clearance is the deadline that matters. It creates a narrow window for new projects to enter the queue with a realistic path to contributing in the 2030-2035 period. Treat that window as a deadline, not a reference point.

Ireland shows where this trajectory leads. Data centres reached 23% of national electricity consumption in 2025, prompting authorities to pause new connections in the Dublin region until at least 2028. Synapse expects PJM regional demand to nearly double by 2040, driven almost entirely by data centres, and Pennsylvania’s confirmed 6,517 MW shortfall in the 2027/2028 delivery year suggests the US is approaching a comparable inflection point. The open question is which state acts next.

This article is for informational purposes only and should not be considered financial advice. Investors should conduct their own research and consult with financial professionals before making investment decisions. Past performance does not guarantee future results, and financial projections are subject to market conditions and various risk factors. Forward-looking statements are speculative and subject to change based on market developments.